asl_recognition.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# American Sign Language (ASL) Image Recognition"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Introduction"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Loading in dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "ename": "ModuleNotFoundError",
     "evalue": "No module named 'tensorflow'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mModuleNotFoundError\u001b[0m                       Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-5-66345e624eb3>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      6\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0msklearn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdatasets\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mload_files\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;31m# from keras.utils import np_utils\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 8\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0mkeras\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpreprocessing\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mimage\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      9\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtqdm\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mtqdm\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     10\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/keras/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0m__future__\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mabsolute_import\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      2\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 3\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mutils\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      4\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mactivations\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mapplications\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/keras/utils/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mdata_utils\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mio_utils\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 6\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mconv_utils\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      7\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;31m# Globally-importable utils.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/keras/utils/conv_utils.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0msix\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmoves\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      8\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mnumpy\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 9\u001b[0;31m \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0;34m.\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mbackend\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mK\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     10\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     11\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/keras/backend/__init__.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     87\u001b[0m \u001b[0;32melif\u001b[0m \u001b[0m_BACKEND\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m'tensorflow'\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     88\u001b[0m     \u001b[0msys\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mstderr\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mwrite\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Using TensorFlow backend.\\n'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 89\u001b[0;31m     \u001b[0;32mfrom\u001b[0m \u001b[0;34m.\u001b[0m\u001b[0mtensorflow_backend\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     90\u001b[0m \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     91\u001b[0m     \u001b[0;31m# Try and load external backend.\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/anaconda3/lib/python3.7/site-packages/keras/backend/tensorflow_backend.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0m__future__\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mprint_function\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0;32mimport\u001b[0m \u001b[0mtensorflow\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtf\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      6\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtensorflow\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpython\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mframework\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mops\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0mtf_ops\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      7\u001b[0m \u001b[0;32mfrom\u001b[0m \u001b[0mtensorflow\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mpython\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtraining\u001b[0m \u001b[0;32mimport\u001b[0m \u001b[0mmoving_averages\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mModuleNotFoundError\u001b[0m: No module named 'tensorflow'"
     ]
    }
   ],
   "source": [
    "import re\n",
    "import os\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "from PIL import Image, ImageFile  \n",
    "from sklearn.datasets import load_files\n",
    "from keras.utils import np_utils\n",
    "from keras.preprocessing import image                  \n",
    "from tqdm import tqdm\n",
    "\n",
    "# For compiling model\n",
    "from keras.callbacks import ModelCheckpoint   \n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Conv2D, GlobalAveragePooling2D, MaxPooling2D\n",
    "\n",
    "# For transfer learning\n",
    "import scipy\n",
    "import keras.applications.vgg16 as vgg16\n",
    "import keras.applications.vgg19 as vgg19\n",
    "import keras.applications.resnet50 as resnet50\n",
    "\n",
    "# For confusion matrix calculation/visualization\n",
    "import tensorflow as tf\n",
    "import seaborn as sn\n",
    "import pandas as pd\n",
    "\n",
    "models_dir = 'saved_models'\n",
    "data_dir = 'data'\n",
    "bottleneck_feat_dir = 'bottleneck_features'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dividing data into training, validation, and testing sets\n",
    "\n",
    "Now that preprocessing the images is completed (see `data_preprocessing.ipynb` notebook), the full dataset will be split into training, validation, and testing sets. The testing set will be all the images from one subject to mirror the \"Spelling It Out\" paper's method so the benchmark model can be compared. The rest of the images will be randomly split; 80% of images for training, 20% of the images for validation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_dataset(path, n_categories=24):\n",
    "    data = load_files(path)\n",
    "    image_files = np.array(data['filenames'])\n",
    "    # Hot encode categories to matrix\n",
    "    image_targets = np_utils.to_categorical(np.array(data['target']), n_categories)\n",
    "    return image_files, image_targets\n",
    "\n",
    "def move_data_by_category(container_dir, regex_file_format='.*png'):\n",
    "    '''Move data into a directory based on category'''\n",
    "    # Still check if files are images\n",
    "    file_list = [x for x in os.listdir(container_dir) if re.search(regex_file_format, x)]\n",
    "    # Get numerical string (note that 1 digits are represented w/ 2 digits) \n",
    "    letters = {x.split('_')[0] for x in file_list}\n",
    "    \n",
    "    for letter in letters:\n",
    "        # Only images that match letter\n",
    "        images_with_letter = [filename for filename in file_list if filename.split('_')[0] == letter]\n",
    "        # Add images to sub directory\n",
    "        new_categ_path = os.path.join(container_dir, letter)\n",
    "        if not os.path.exists(new_categ_path):\n",
    "            os.makedirs(new_categ_path)\n",
    "#         print('Created {new_categ_path} dir with {len(images_with_letter)} items')\n",
    "        for img_filename in images_with_letter:\n",
    "            path = os.path.join(container_dir, img_filename)\n",
    "            new_path = os.path.join(new_categ_path, img_filename)            \n",
    "            os.rename(path, new_path)\n",
    "    # TODO: Check if any files were skipped (improperly named?)\n",
    "        \n",
    "\n",
    "def get_testing_data(data_dir, subject_num='4'):\n",
    "    '''Get all data/images pertaining to one subject'''\n",
    "    # Only search in directory for images with that subject\n",
    "    file_list = [x for x in os.listdir(data_dir) if re.search('\\d+_{}_\\d*\\.png'.format(subject_num), x)]\n",
    "    \n",
    "    # Make a new testing data directory if doesn't exist\n",
    "    testing_dir = os.path.join(data_dir, 'test')\n",
    "    if not os.path.exists(testing_dir):\n",
    "        os.makedirs(testing_dir)\n",
    "        \n",
    "    # Move images of particular subject into testing directory\n",
    "    for image_filename in file_list:\n",
    "        # file is **_n_****.png where n is an integer representing a subject\n",
    "        _, subject, _ = image_filename.split('_')\n",
    "        # Move file into testing directory\n",
    "        path = os.path.join(data_dir, image_filename)\n",
    "        new_path = os.path.join(testing_dir, image_filename)\n",
    "        os.rename(path, new_path)\n",
    "        \n",
    "    # Move each image file's numerical str representing letters found in testing into own category directory\n",
    "    move_data_by_category(testing_dir)\n",
    "    \n",
    "    return load_dataset(testing_dir)\n",
    "\n",
    "\n",
    "def get_training_validation_data(data_dir, ratio=0.8):\n",
    "    '''Randomly split data into training and validation sets'''\n",
    "    # Only search in directory for images\n",
    "    file_list = [x for x in os.listdir(data_dir) if re.search('.*png', x)]\n",
    "    \n",
    "    # Make a new training & validation data directory if doesn't exist\n",
    "    train_dir = os.path.join(data_dir, 'train')\n",
    "    valid_dir = os.path.join(data_dir, 'valid')\n",
    "    if not os.path.exists(train_dir):\n",
    "        os.makedirs(train_dir)\n",
    "    if not os.path.exists(valid_dir):\n",
    "        os.makedirs(valid_dir)\n",
    "        \n",
    "    # Randomly split file list into training and vaidation sets\n",
    "    np.random.shuffle(file_list)\n",
    "    split_int = int(ratio * len(file_list))\n",
    "    train_list = file_list[:split_int]\n",
    "    valid_list = file_list[split_int:]\n",
    "    \n",
    "    # Move images of particular subject into testing directory\n",
    "    for filenames, new_dir in [(train_list, train_dir), (valid_list, valid_dir)]:\n",
    "        for image_filename in filenames:\n",
    "            # Move file into testing directory\n",
    "            path = os.path.join(data_dir, image_filename)\n",
    "            new_path = os.path.join(new_dir, image_filename)\n",
    "            os.rename(path, new_path)\n",
    "\n",
    "        # Move each image file's numerical str representing letters found in testing into own category directory\n",
    "        move_data_by_category(new_dir)\n",
    "    \n",
    "    return (load_dataset(train_dir), load_dataset(valid_dir))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Manually set if the data is RGB\n",
    "is_rgb = True\n",
    "\n",
    "test_data, test_targets = get_testing_data(data_dir)\n",
    "train, valid = get_training_validation_data(data_dir)\n",
    "# Separated data and its targets\n",
    "train_data, train_targets = train\n",
    "valid_data, valid_targets = valid"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Display some of the images"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "np.random.seed(8675309)\n",
    "%matplotlib inline\n",
    "\n",
    "# Display image previews below\n",
    "plt.figure(figsize=(20,55))\n",
    "columns = 8\n",
    "n = 1\n",
    "\n",
    "# Randomly choose images to display (with label)\n",
    "for image_path in np.random.choice(train_data, 24, replace=False):\n",
    "    img = Image.open(image_path)\n",
    "    plt.subplot(20, columns, n)\n",
    "    n+=1\n",
    "    plt.imshow(img)\n",
    "    letter = image_path.split('/')[-1][:2]\n",
    "    letter = chr(int(letter)+65)\n",
    "    plt.title(letter)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preparing the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preprocessing data fro Keras (TensorFlow backend)\n",
    "def path_to_tensor(img_path):\n",
    "    # Loads image as PIL.Image.Image type\n",
    "    img = image.load_img(img_path, target_size=(224, 224), grayscale=False)\n",
    "    # Convert PIL.Image.Image type to 3D tensor with shape (224, 224, 1)\n",
    "    x = image.img_to_array(img)\n",
    "    # convert 3D tensor to 4D tensor with shape (1, 224, 224, 1) and return 4D tensor\n",
    "    return np.expand_dims(x, axis=0)\n",
    "\n",
    "def paths_to_tensor(img_paths):\n",
    "    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]\n",
    "    return np.vstack(list_of_tensors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ImageFile.LOAD_TRUNCATED_IMAGES = True                 \n",
    "\n",
    "# pre-process the data for Keras\n",
    "valid_tensors = paths_to_tensor(valid_data).astype('float32')\n",
    "test_tensors = paths_to_tensor(test_data).astype('float32')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Prepare training data separately since this is usually large\n",
    "train_tensors = paths_to_tensor(train_data).astype('float32')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Basic CNN model from scratch"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building basic model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.layers import Conv2D, MaxPooling2D, GlobalAveragePooling2D\n",
    "from keras.layers import Dropout, Flatten, Dense\n",
    "from keras.models import Sequential\n",
    "\n",
    "model = Sequential()\n",
    "#Convo 224, 224, 1\n",
    "model.add(Conv2D(filters=16, kernel_size=2, padding='same', activation='relu', \n",
    "                        input_shape=(224, 224, 1)))\n",
    "#\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "#\n",
    "model.add(Conv2D(filters=32, kernel_size=2, padding='same', activation='relu'))\n",
    "#\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "#\n",
    "model.add(Conv2D(filters=64, kernel_size=2, padding='same', activation='relu'))\n",
    "#\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "#\n",
    "model.add(GlobalAveragePooling2D())\n",
    "#Dense; 24 for each handshape \n",
    "model.add(Dense(24, activation='softmax'))\n",
    "\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.compile(optimizer='rmsprop', loss='categorical_crossentropy', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training basic model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.callbacks import ModelCheckpoint  \n",
    "\n",
    "epochs = 16\n",
    "\n",
    "# Create a saved models directory\n",
    "if not os.path.exists(models_dir):\n",
    "    os.makedirs(models_dir)\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=f'{models_dir}/weights.best.from_scratch.hdf5', \n",
    "                               verbose=1, save_best_only=True)\n",
    "\n",
    "model.fit(train_tensors, train_targets, \n",
    "          validation_data=(valid_tensors, valid_targets),\n",
    "          epochs=epochs, batch_size=20, callbacks=[checkpointer], verbose=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating basic model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "handshape_predictions = [np.argmax(model.predict(np.expand_dims(tensor, axis=0))) for tensor in test_tensors]\n",
    "\n",
    "# report test accuracy\n",
    "test_accuracy = 100*np.sum(np.array(handshape_predictions)==np.argmax(test_targets, axis=1))/len(handshape_predictions)\n",
    "print('Test accuracy: %.4f%%' % test_accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transfer Learning with VGG16"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preparing the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define image size to use for given model\n",
    "n_pixels = 80\n",
    "# Define a model name for records\n",
    "model_name ='rgbModel_{}px'.format(n_pixels) if is_rgb else 'model_{}px'.format(n_pixels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preprocessing data fro Keras (TensorFlow backend)\n",
    "def path_to_tensor(img_path):\n",
    "    # Loads image as PIL.Image.Image type\n",
    "    img = image.load_img(img_path, target_size=(n_pixels, n_pixels), grayscale=False)\n",
    "    # Convert PIL.Image.Image type to 3D tensor with shape (n, n, 3)\n",
    "    x = image.img_to_array(img)\n",
    "    # convert 3D tensor to 4D tensor with shape (1, n, n, 3) and return 4D tensor\n",
    "    return np.expand_dims(x, axis=0)\n",
    "\n",
    "def paths_to_tensor(img_paths):\n",
    "    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]\n",
    "    return np.vstack(list_of_tensors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ImageFile.LOAD_TRUNCATED_IMAGES = True                 \n",
    "# Pre-process the data for Keras\n",
    "# Prepare training data separately since this is usually large\n",
    "\n",
    "print('Preparing to create valid tensors')\n",
    "valid_tensors = paths_to_tensor(valid_data).astype('float32')\n",
    "print('valid tensors prepared')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Preparing to create test tensors')\n",
    "test_tensors = paths_to_tensor(test_data).astype('float32')\n",
    "print('test tensors prepared')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Preparing to create train tensors')\n",
    "train_tensors = paths_to_tensor(train_data).astype('float32')\n",
    "print('train tensors prepared')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Training Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Name & path of bottleneck features\n",
    "feat_train_name = 'vgg16_{}_features_train'.format(model_name)\n",
    "feat_train_path = '{}/{}.npz'.format(bottleneck_feat_dir,feat_train_name)\n",
    "\n",
    "# Load the data\n",
    "targets = np.squeeze(train_targets)\n",
    "print('data loaded')\n",
    "\n",
    "# Load vgg16 model + remove final classification layers\n",
    "model = vgg16.VGG16(weights='imagenet', include_top=False, input_shape=(n_pixels, n_pixels, 3), classes=24)\n",
    "print('model loaded')\n",
    "\n",
    "# Obtain bottleneck features (train)\n",
    "if os.path.exists(feat_train_path):\n",
    "    print('bottleneck features detected (train)')\n",
    "    features = np.load(feat_train_path)['features']\n",
    "else:\n",
    "    print('bottleneck features file not detected (train)')\n",
    "    \n",
    "    print('calculating now ...')\n",
    "    # Pre-process the train data\n",
    "    big_x_train = np.array([scipy.misc.imresize(train_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                            for i in range(0, len(train_tensors))]).astype('float32')\n",
    "    vgg16_input_train = vgg16.preprocess_input(big_x_train)\n",
    "    print('train data preprocessed')\n",
    "    \n",
    "    # Extract, process, and save bottleneck features\n",
    "    features = model.predict(vgg16_input_train)\n",
    "    features = np.squeeze(features)\n",
    "    np.savez(feat_train_path, features=features)\n",
    "\n",
    "print('bottleneck features saved (train)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Validation Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Name & path of bottleneck features\n",
    "feat_valid_name = 'vgg16_{}_features_valid'.format(model_name)\n",
    "feat_valid_path = '{}/{}.npz'.format(bottleneck_feat_dir,feat_valid_name)\n",
    "\n",
    "# Obtain bottleneck features (valid)\n",
    "if os.path.exists(feat_valid_path):\n",
    "    print('bottleneck features detected (valid)')\n",
    "    features_valid = np.load(feat_valid_path)['features_test']\n",
    "else:\n",
    "    # Calculating for validation set\n",
    "    print('bottleneck features file not detected (valid)')\n",
    "    print('calculating now ...')\n",
    "    # Pre-process the valid data\n",
    "    big_x_valid = np.array([scipy.misc.imresize(valid_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                       for i in range(0, len(valid_tensors))]).astype('float32')\n",
    "\n",
    "    vgg16_input_valid = vgg16.preprocess_input(big_x_valid)\n",
    "    # Extract, process, and save bottleneck features (valid)\n",
    "    features_valid = model.predict(vgg16_input_valid)\n",
    "    features_valid = np.squeeze(features_valid)\n",
    "    np.savez(feat_valid_path, features_test=features_valid)\n",
    "print('bottleneck features saved (valid)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Test Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Name & path of bottleneck features\n",
    "feat_test_name = 'vgg16_{}_features_test'.format(model_name)\n",
    "feat_test_path = '{}/{}.npz'.format(bottleneck_feat_dir,feat_test_name)\n",
    "\n",
    "# Obtain bottleneck features (test)\n",
    "if os.path.exists(feat_test_path):\n",
    "    print('bottleneck features detected (test)')\n",
    "    features_test = np.load(feat_test_path)['features_test']\n",
    "else:\n",
    "    # Calculating for test set\n",
    "    print('bottleneck features file not detected (test)')\n",
    "    print('calculating now ...')\n",
    "    # pre-process the test data\n",
    "    big_x_test = np.array([scipy.misc.imresize(test_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                       for i in range(0, len(test_tensors))]).astype('float32')\n",
    "\n",
    "    vgg16_input_test = vgg16.preprocess_input(big_x_test)\n",
    "    # Extract, process, and save bottleneck features (test)\n",
    "    features_test = model.predict(vgg16_input_test)\n",
    "    features_test = np.squeeze(features_test)\n",
    "    np.savez(feat_test_path, features_test=features_test)\n",
    "print('bottleneck features saved (test)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "features.shape\n",
    "type(features)\n",
    "features_ = features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 3 CNN Layers\n",
    "model = Sequential()\n",
    "\n",
    "model.add(Conv2D(filters=256, kernel_size=2, input_shape=features.shape[1:]))\n",
    "model.add(Dropout(0.4))\n",
    "\n",
    "model.add(Conv2D(filters=128, kernel_size=1))\n",
    "\n",
    "model.add(Conv2D(filters=32, kernel_size=1))\n",
    "model.add(Dropout(0.3))\n",
    "\n",
    "model.add(GlobalAveragePooling2D())\n",
    "model.add(Dropout(0.3))\n",
    "\n",
    "model.add(Dense(24, activation='softmax'))\n",
    "model.summary()\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 50\n",
    "batchsize = 500\n",
    "model_weights_path = '{}/vgg16_{}.best_batch{}_epoch{}.hdf5'.format(models_dir,model_name,batchsize,epochs)\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=model_weights_path, \n",
    "                               verbose=1, save_best_only=True)\n",
    "\n",
    "model.fit(features, targets, batch_size=batchsize, epochs=epochs,\n",
    "          validation_data=(features_valid, valid_targets),\n",
    "          callbacks=[checkpointer],\n",
    "          verbose=2, shuffle=True)          "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the weights that yielded the best validation accuracy\n",
    "model.load_weights(model_weights_path)\n",
    "\n",
    "# Evaluate test accuracy\n",
    "score = model.evaluate(features_test, test_targets, verbose=0)\n",
    "accuracy = 100*score[1]\n",
    "\n",
    "# Print test accuracy\n",
    "print('Test accuracy: %.4f%%' % accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transfer Learning with VGG19"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preparing the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define image size to use for given model\n",
    "n_pixels = 160\n",
    "# Define a model name for recorcs\n",
    "model_name = 'model_{}px'.format(n_pixels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preprocessing data fro Keras (TensorFlow backend)\n",
    "def path_to_tensor(img_path):\n",
    "    # Loads image as PIL.Image.Image type\n",
    "    img = image.load_img(img_path, target_size=(n_pixels, n_pixels), grayscale=False)\n",
    "    # Convert PIL.Image.Image type to 3D tensor with shape (n, n, 3)\n",
    "    x = image.img_to_array(img)\n",
    "    # convert 3D tensor to 4D tensor with shape (1, n, n, 3) and return 4D tensor\n",
    "    return np.expand_dims(x, axis=0)\n",
    "\n",
    "def paths_to_tensor(img_paths):\n",
    "    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]\n",
    "    return np.vstack(list_of_tensors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "ImageFile.LOAD_TRUNCATED_IMAGES = True                 \n",
    "# Pre-process the data for Keras\n",
    "# Prepare training data separately since this is usually large\n",
    "\n",
    "print('Preparing to create valid tensors')\n",
    "valid_tensors = paths_to_tensor(valid_data).astype('float32')\n",
    "print('valid tensors prepared')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Preparing to create test tensors')\n",
    "test_tensors = paths_to_tensor(test_data).astype('float32')\n",
    "print('test tensors prepared')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Preparing to create train tensors')\n",
    "train_tensors = paths_to_tensor(train_data).astype('float32')\n",
    "print('train tensors prepared')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Training Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the data\n",
    "targets = np.squeeze(train_targets)\n",
    "print('data loaded')\n",
    "\n",
    "# Load vgg19 model + remove final classification layers\n",
    "model = vgg19.VGG19(weights='imagenet', include_top=False, input_shape=(n_pixels, n_pixels, 3), classes=24)\n",
    "print('model loaded')\n",
    "\n",
    "# Obtain bottleneck features (train)\n",
    "if os.path.exists('vgg19_features_train.npz'):\n",
    "    print('bottleneck features detected (train)')\n",
    "    features = np.load('vgg19_features_train.npz')['features']\n",
    "else:\n",
    "    print('bottleneck features file not detected (train)')\n",
    "    \n",
    "    print('calculating now ...')\n",
    "    # Pre-process the train data\n",
    "    big_x_train = np.array([scipy.misc.imresize(train_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                            for i in range(0, len(train_tensors))]).astype('float32')\n",
    "    vgg19_input_train = vgg19.preprocess_input(big_x_train)\n",
    "    print('train data preprocessed')\n",
    "    \n",
    "    # Extract, process, and save bottleneck features\n",
    "    features = model.predict(vgg19_input_train)\n",
    "    features = np.squeeze(features)\n",
    "    np.savez('vgg19_features_train', features=features)\n",
    "\n",
    "print('bottleneck features saved (train)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Validation Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Obtain bottleneck features (valid)\n",
    "if os.path.exists('vgg19_features_valid.npz'):\n",
    "    print('bottleneck features detected (valid)')\n",
    "    features_valid = np.load('vgg19_features_valid.npz')['features_test']\n",
    "else:\n",
    "    # Calculating for validation set\n",
    "    print('bottleneck features file not detected (valid)')\n",
    "    print('calculating now ...')\n",
    "    # Pre-process the valid data\n",
    "    big_x_valid = np.array([scipy.misc.imresize(valid_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                       for i in range(0, len(valid_tensors))]).astype('float32')\n",
    "\n",
    "    vgg19_input_valid = vgg19.preprocess_input(big_x_valid)\n",
    "    # Extract, process, and save bottleneck features (valid)\n",
    "    features_valid = model.predict(vgg19_input_valid)\n",
    "    features_valid = np.squeeze(features_valid)\n",
    "    np.savez('vgg19_features_valid', features_test=features_valid)\n",
    "print('bottleneck features saved (valid)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Test Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# Obtain bottleneck features (test)\n",
    "if os.path.exists('vgg19_features_test.npz'):\n",
    "    print('bottleneck features detected (test)')\n",
    "    features_test = np.load('vgg19_features_test.npz')['features_test']\n",
    "else:\n",
    "    # Calculating for test set\n",
    "    print('bottleneck features file not detected (test)')\n",
    "    print('calculating now ...')\n",
    "    # pre-process the test data\n",
    "    big_x_test = np.array([scipy.misc.imresize(test_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                       for i in range(0, len(test_tensors))]).astype('float32')\n",
    "\n",
    "    vgg19_input_test = vgg19.preprocess_input(big_x_test)\n",
    "    # Extract, process, and save bottleneck features (test)\n",
    "    features_test = model.predict(vgg19_input_test)\n",
    "    features_test = np.squeeze(features_test)\n",
    "    np.savez('vgg19_features_test', features_test=features_test)\n",
    "print('bottleneck features saved (test)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Shallow NN\n",
    "from keras.callbacks import ModelCheckpoint   \n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Dropout, Conv2D, GlobalAveragePooling2D\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Conv2D(filters=256, kernel_size=2, input_shape=features.shape[1:]))\n",
    "model.add(Dropout(0.4))\n",
    "model.add(GlobalAveragePooling2D())\n",
    "model.add(Dropout(0.3))\n",
    "model.add(Dense(24, activation='softmax'))\n",
    "model.summary()\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "epochs = 256\n",
    "batchsize = 512\n",
    "model_weights_path = f'{models_dir}/vgg19_new_{model_name}.best_batch{batchsize}_epoch{epochs}.hdf5'\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=model_weights_path, \n",
    "                               verbose=1, save_best_only=True)\n",
    "\n",
    "model.fit(features, targets, batch_size=batchsize, epochs=epochs,\n",
    "          validation_data=(features_valid, valid_targets),\n",
    "          callbacks=[checkpointer],\n",
    "          verbose=2, shuffle=True)          "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the weights that yielded the best validation accuracy\n",
    "model.load_weights(model_weights_path)\n",
    "\n",
    "# Evaluate test accuracy\n",
    "score = model.evaluate(features_test, test_targets, verbose=0)\n",
    "accuracy = 100*score[1]\n",
    "\n",
    "# Print test accuracy\n",
    "print('Test accuracy: %.4f%%' % accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transfer Learning with ResNet50"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preparing the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define image size to use for given model\n",
    "n_pixels = 197\n",
    "# Define a model name for records\n",
    "model_name ='rgbModel_{}px'.format(n_pixels) if is_rgb else 'model_{}px'.format(n_pixels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Preprocessing data fro Keras (TensorFlow backend)\n",
    "def path_to_tensor(img_path):\n",
    "    # Loads image as PIL.Image.Image type\n",
    "    img = image.load_img(img_path, target_size=(n_pixels, n_pixels), grayscale=False)\n",
    "    # Convert PIL.Image.Image type to 3D tensor with shape (n, n, 3)\n",
    "    x = image.img_to_array(img)\n",
    "    # convert 3D tensor to 4D tensor with shape (1, n, n, 3) and return 4D tensor\n",
    "    return np.expand_dims(x, axis=0)\n",
    "\n",
    "def paths_to_tensor(img_paths):\n",
    "    list_of_tensors = [path_to_tensor(img_path) for img_path in tqdm(img_paths)]\n",
    "    return np.vstack(list_of_tensors)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  1%|          | 69/10599 [00:00<00:15, 686.63it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Preparing to create valid tensors\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 10599/10599 [00:16<00:00, 658.26it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "valid tensors prepared\n"
     ]
    }
   ],
   "source": [
    "ImageFile.LOAD_TRUNCATED_IMAGES = True                 \n",
    "# Pre-process the data for Keras\n",
    "# Prepare training data separately since this is usually large\n",
    "\n",
    "print('Preparing to create valid tensors')\n",
    "valid_tensors = paths_to_tensor(valid_data).astype('float32')\n",
    "print('valid tensors prepared')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|          | 58/12782 [00:00<00:22, 577.28it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Preparing to create test tensors\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 12782/12782 [00:21<00:00, 594.40it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "test tensors prepared\n"
     ]
    }
   ],
   "source": [
    "print('Preparing to create test tensors')\n",
    "test_tensors = paths_to_tensor(test_data).astype('float32')\n",
    "print('test tensors prepared')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Preparing to create train tensors\n",
      "train tensors prepared\n"
     ]
    }
   ],
   "source": [
    "print('Preparing to create train tensors')\n",
    "train_tensors = paths_to_tensor(train_data).astype('float32')\n",
    "print('train tensors prepared')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Training Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "data loaded\n",
      "model loaded\n",
      "bottleneck features detected (train)\n",
      "bottleneck features saved (train)\n"
     ]
    }
   ],
   "source": [
    "# Name & path of bottleneck features\n",
    "feat_train_name = 'resnet50_{}_features_train'.format(model_name)\n",
    "feat_train_path = '{}/{}.npz'.format(bottleneck_feat_dir,feat_train_name)\n",
    "\n",
    "# Load the data\n",
    "targets = np.squeeze(train_targets)\n",
    "print('data loaded')\n",
    "\n",
    "# Load resnet50 model + remove final classification layers\n",
    "model = resnet50.ResNet50(weights='imagenet', include_top=False, input_shape=(n_pixels, n_pixels, 3), classes=24)\n",
    "print('model loaded')\n",
    "\n",
    "# Obtain bottleneck features (train)\n",
    "if os.path.exists(feat_train_path):\n",
    "    print('bottleneck features detected (train)')\n",
    "    features = np.load(feat_train_path)['features']\n",
    "else:\n",
    "    print('bottleneck features file not detected (train)')\n",
    "    \n",
    "    print('calculating now ...')\n",
    "    # Pre-process the train data\n",
    "    big_x_train = np.array([scipy.misc.imresize(train_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                            for i in range(0, len(train_tensors))]).astype('float32')\n",
    "    resnet50_input_train = resnet50.preprocess_input(big_x_train)\n",
    "    print('train data preprocessed')\n",
    "    \n",
    "    # Extract, process, and save bottleneck features\n",
    "    features = model.predict(resnet50_input_train)\n",
    "#     features = np.squeeze(features)\n",
    "    np.savez(feat_train_path, features=features)\n",
    "\n",
    "print('bottleneck features saved (train)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Validation Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bottleneck features detected (valid)\n",
      "bottleneck features saved (valid)\n"
     ]
    }
   ],
   "source": [
    "# Name & path of bottleneck features\n",
    "feat_valid_name = 'resnet50_{}_features_valid'.format(model_name)\n",
    "feat_valid_path = '{}/{}.npz'.format(bottleneck_feat_dir,feat_valid_name)\n",
    "\n",
    "# Obtain bottleneck features (valid)\n",
    "if os.path.exists(feat_valid_path):\n",
    "    print('bottleneck features detected (valid)')\n",
    "    features_valid = np.load(feat_valid_path)['features_test']\n",
    "else:\n",
    "    # Calculating for validation set\n",
    "    print('bottleneck features file not detected (valid)')\n",
    "    print('calculating now ...')\n",
    "    # Pre-process the valid data\n",
    "    big_x_valid = np.array([scipy.misc.imresize(valid_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                       for i in range(0, len(valid_tensors))]).astype('float32')\n",
    "\n",
    "    resnet50_input_valid = resnet50.preprocess_input(big_x_valid)\n",
    "    # Extract, process, and save bottleneck features (valid)\n",
    "    features_valid = model.predict(resnet50_input_valid)\n",
    "#     features_valid = np.squeeze(features_valid)\n",
    "    np.savez(feat_valid_path, features_test=features_valid)\n",
    "print('bottleneck features saved (valid)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Extract Bottleneck Features for Test Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "bottleneck features detected (test)\n",
      "bottleneck features saved (test)\n"
     ]
    }
   ],
   "source": [
    "# Name & path of bottleneck features\n",
    "feat_test_name = 'resnet50_{}_features_test'.format(model_name)\n",
    "feat_test_path = '{}/{}.npz'.format(bottleneck_feat_dir,feat_test_name)\n",
    "\n",
    "# Obtain bottleneck features (test)\n",
    "if os.path.exists(feat_test_path):\n",
    "    print('bottleneck features detected (test)')\n",
    "    features_test = np.load(feat_test_path)['features_test']\n",
    "else:\n",
    "    # Calculating for test set\n",
    "    print('bottleneck features file not detected (test)')\n",
    "    print('calculating now ...')\n",
    "    # pre-process the test data\n",
    "    big_x_test = np.array([scipy.misc.imresize(test_tensors[i], (n_pixels, n_pixels, 3)) \n",
    "                       for i in range(0, len(test_tensors))]).astype('float32')\n",
    "\n",
    "    resnet50_input_test = resnet50.preprocess_input(big_x_test)\n",
    "    # Extract, process, and save bottleneck features (test)\n",
    "    features_test = model.predict(resnet50_input_test)\n",
    "#     features_test = np.squeeze(features_test)\n",
    "    np.savez(feat_test_path, features_test=features_test)\n",
    "print('bottleneck features saved (test)')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Building model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 1, 1, 512)         1049088   \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 1, 1, 512)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 1, 1, 256)         131328    \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 1, 1, 256)         0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 1, 1, 256)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 1, 1, 128)         32896     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 1, 1, 128)         0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 1, 1, 128)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 1, 1, 128)         16512     \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_1 ( (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 24)                3096      \n",
      "=================================================================\n",
      "Total params: 1,232,920\n",
      "Trainable params: 1,232,920\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# 3 CNN Layers\n",
    "from keras.layers import MaxPooling2D\n",
    "\n",
    "model = Sequential()\n",
    "\n",
    "model.add(Conv2D(filters=512, kernel_size=1, input_shape=features.shape[1:]))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=256, kernel_size=1))\n",
    "model.add(MaxPooling2D(pool_size=1))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=128, kernel_size=1))\n",
    "model.add(MaxPooling2D(pool_size=1))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=128, kernel_size=1))\n",
    "\n",
    "model.add(GlobalAveragePooling2D())\n",
    "model.add(Dropout(0.3))\n",
    "\n",
    "model.add(Dense(24, activation='softmax'))\n",
    "model.summary()\n",
    "\n",
    "model.compile(loss='categorical_crossentropy', optimizer='rmsprop', metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 42393 samples, validate on 10599 samples\n",
      "Epoch 1/1536\n",
      "Epoch 00000: val_loss improved from inf to 0.05668, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0279 - acc: 0.9910 - val_loss: 0.0567 - val_acc: 0.9825\n",
      "Epoch 2/1536\n",
      "Epoch 00001: val_loss did not improve\n",
      "0s - loss: 0.0281 - acc: 0.9908 - val_loss: 0.0627 - val_acc: 0.9791\n",
      "Epoch 3/1536\n",
      "Epoch 00002: val_loss did not improve\n",
      "0s - loss: 0.0797 - acc: 0.9749 - val_loss: 0.5369 - val_acc: 0.8987\n",
      "Epoch 4/1536\n",
      "Epoch 00003: val_loss did not improve\n",
      "0s - loss: 0.6408 - acc: 0.8752 - val_loss: 0.5027 - val_acc: 0.8886\n",
      "Epoch 5/1536\n",
      "Epoch 00004: val_loss did not improve\n",
      "0s - loss: 0.2616 - acc: 0.9353 - val_loss: 0.0622 - val_acc: 0.9792\n",
      "Epoch 6/1536\n",
      "Epoch 00005: val_loss did not improve\n",
      "0s - loss: 0.0330 - acc: 0.9889 - val_loss: 0.0580 - val_acc: 0.9808\n",
      "Epoch 7/1536\n",
      "Epoch 00006: val_loss did not improve\n",
      "0s - loss: 0.0278 - acc: 0.9913 - val_loss: 0.0599 - val_acc: 0.9802\n",
      "Epoch 8/1536\n",
      "Epoch 00007: val_loss did not improve\n",
      "0s - loss: 0.0301 - acc: 0.9899 - val_loss: 0.0575 - val_acc: 0.9811\n",
      "Epoch 9/1536\n",
      "Epoch 00008: val_loss did not improve\n",
      "0s - loss: 0.0317 - acc: 0.9888 - val_loss: 0.0835 - val_acc: 0.9725\n",
      "Epoch 10/1536\n",
      "Epoch 00009: val_loss did not improve\n",
      "0s - loss: 0.8056 - acc: 0.8745 - val_loss: 0.9524 - val_acc: 0.8303\n",
      "Epoch 11/1536\n",
      "Epoch 00010: val_loss did not improve\n",
      "0s - loss: 0.9819 - acc: 0.8248 - val_loss: 0.0780 - val_acc: 0.9754\n",
      "Epoch 12/1536\n",
      "Epoch 00011: val_loss did not improve\n",
      "0s - loss: 0.0460 - acc: 0.9837 - val_loss: 0.0666 - val_acc: 0.9795\n",
      "Epoch 13/1536\n",
      "Epoch 00012: val_loss did not improve\n",
      "0s - loss: 0.0345 - acc: 0.9879 - val_loss: 0.0655 - val_acc: 0.9794\n",
      "Epoch 14/1536\n",
      "Epoch 00013: val_loss did not improve\n",
      "0s - loss: 0.0338 - acc: 0.9884 - val_loss: 0.0616 - val_acc: 0.9820\n",
      "Epoch 15/1536\n",
      "Epoch 00014: val_loss did not improve\n",
      "0s - loss: 0.0315 - acc: 0.9892 - val_loss: 0.0647 - val_acc: 0.9808\n",
      "Epoch 16/1536\n",
      "Epoch 00015: val_loss did not improve\n",
      "0s - loss: 0.0297 - acc: 0.9902 - val_loss: 0.0866 - val_acc: 0.9744\n",
      "Epoch 17/1536\n",
      "Epoch 00016: val_loss did not improve\n",
      "0s - loss: 0.2558 - acc: 0.9335 - val_loss: 0.7897 - val_acc: 0.8353\n",
      "Epoch 18/1536\n",
      "Epoch 00017: val_loss did not improve\n",
      "0s - loss: 0.7957 - acc: 0.8436 - val_loss: 0.1211 - val_acc: 0.9634\n",
      "Epoch 19/1536\n",
      "Epoch 00018: val_loss did not improve\n",
      "0s - loss: 0.0464 - acc: 0.9845 - val_loss: 0.0648 - val_acc: 0.9800\n",
      "Epoch 20/1536\n",
      "Epoch 00019: val_loss did not improve\n",
      "0s - loss: 0.0327 - acc: 0.9885 - val_loss: 0.0602 - val_acc: 0.9801\n",
      "Epoch 21/1536\n",
      "Epoch 00020: val_loss did not improve\n",
      "0s - loss: 0.0330 - acc: 0.9886 - val_loss: 0.0641 - val_acc: 0.9785\n",
      "Epoch 22/1536\n",
      "Epoch 00021: val_loss did not improve\n",
      "0s - loss: 0.0330 - acc: 0.9883 - val_loss: 0.0731 - val_acc: 0.9762\n",
      "Epoch 23/1536\n",
      "Epoch 00022: val_loss did not improve\n",
      "0s - loss: 0.3287 - acc: 0.9330 - val_loss: 1.4661 - val_acc: 0.7657\n",
      "Epoch 24/1536\n",
      "Epoch 00023: val_loss did not improve\n",
      "0s - loss: 0.9616 - acc: 0.8355 - val_loss: 0.0753 - val_acc: 0.9758\n",
      "Epoch 25/1536\n",
      "Epoch 00024: val_loss did not improve\n",
      "0s - loss: 0.0428 - acc: 0.9856 - val_loss: 0.0655 - val_acc: 0.9791\n",
      "Epoch 26/1536\n",
      "Epoch 00025: val_loss did not improve\n",
      "0s - loss: 0.0363 - acc: 0.9874 - val_loss: 0.0618 - val_acc: 0.9811\n",
      "Epoch 27/1536\n",
      "Epoch 00026: val_loss did not improve\n",
      "0s - loss: 0.0341 - acc: 0.9891 - val_loss: 0.0641 - val_acc: 0.9800\n",
      "Epoch 28/1536\n",
      "Epoch 00027: val_loss did not improve\n",
      "0s - loss: 0.0333 - acc: 0.9887 - val_loss: 0.0924 - val_acc: 0.9716\n",
      "Epoch 29/1536\n",
      "Epoch 00028: val_loss did not improve\n",
      "0s - loss: 0.6248 - acc: 0.8928 - val_loss: 2.5458 - val_acc: 0.7112\n",
      "Epoch 30/1536\n",
      "Epoch 00029: val_loss did not improve\n",
      "0s - loss: 0.8544 - acc: 0.8747 - val_loss: 0.0738 - val_acc: 0.9759\n",
      "Epoch 31/1536\n",
      "Epoch 00030: val_loss did not improve\n",
      "0s - loss: 0.0409 - acc: 0.9859 - val_loss: 0.0684 - val_acc: 0.9780\n",
      "Epoch 32/1536\n",
      "Epoch 00031: val_loss did not improve\n",
      "0s - loss: 0.0371 - acc: 0.9879 - val_loss: 0.0685 - val_acc: 0.9770\n",
      "Epoch 33/1536\n",
      "Epoch 00032: val_loss did not improve\n",
      "0s - loss: 0.0334 - acc: 0.9891 - val_loss: 0.0652 - val_acc: 0.9773\n",
      "Epoch 34/1536\n",
      "Epoch 00033: val_loss did not improve\n",
      "0s - loss: 0.0324 - acc: 0.9890 - val_loss: 0.1113 - val_acc: 0.9654\n",
      "Epoch 35/1536\n",
      "Epoch 00034: val_loss did not improve\n",
      "0s - loss: 0.3243 - acc: 0.9291 - val_loss: 0.3936 - val_acc: 0.9030\n",
      "Epoch 36/1536\n",
      "Epoch 00035: val_loss did not improve\n",
      "0s - loss: 0.5218 - acc: 0.8843 - val_loss: 0.1351 - val_acc: 0.9608\n",
      "Epoch 37/1536\n",
      "Epoch 00036: val_loss did not improve\n",
      "0s - loss: 0.0469 - acc: 0.9845 - val_loss: 0.0612 - val_acc: 0.9811\n",
      "Epoch 38/1536\n",
      "Epoch 00037: val_loss did not improve\n",
      "0s - loss: 0.0306 - acc: 0.9901 - val_loss: 0.0591 - val_acc: 0.9813\n",
      "Epoch 39/1536\n",
      "Epoch 00038: val_loss did not improve\n",
      "0s - loss: 0.0281 - acc: 0.9907 - val_loss: 0.0596 - val_acc: 0.9807\n",
      "Epoch 40/1536\n",
      "Epoch 00039: val_loss did not improve\n",
      "0s - loss: 0.0329 - acc: 0.9884 - val_loss: 0.5975 - val_acc: 0.9202\n",
      "Epoch 41/1536\n",
      "Epoch 00040: val_loss did not improve\n",
      "0s - loss: 0.8515 - acc: 0.8610 - val_loss: 0.6463 - val_acc: 0.8554\n",
      "Epoch 42/1536\n",
      "Epoch 00041: val_loss did not improve\n",
      "0s - loss: 0.2738 - acc: 0.9390 - val_loss: 0.0615 - val_acc: 0.9798\n",
      "Epoch 43/1536\n",
      "Epoch 00042: val_loss did not improve\n",
      "0s - loss: 0.0339 - acc: 0.9880 - val_loss: 0.0597 - val_acc: 0.9803\n",
      "Epoch 44/1536\n",
      "Epoch 00043: val_loss did not improve\n",
      "0s - loss: 0.0296 - acc: 0.9902 - val_loss: 0.0632 - val_acc: 0.9796\n",
      "Epoch 45/1536\n",
      "Epoch 00044: val_loss did not improve\n",
      "0s - loss: 0.0308 - acc: 0.9895 - val_loss: 0.0766 - val_acc: 0.9765\n",
      "Epoch 46/1536\n",
      "Epoch 00045: val_loss did not improve\n",
      "0s - loss: 0.1605 - acc: 0.9558 - val_loss: 0.4152 - val_acc: 0.9054\n",
      "Epoch 47/1536\n",
      "Epoch 00046: val_loss did not improve\n",
      "0s - loss: 0.6363 - acc: 0.8718 - val_loss: 0.2123 - val_acc: 0.9412\n",
      "Epoch 48/1536\n",
      "Epoch 00047: val_loss did not improve\n",
      "0s - loss: 0.0884 - acc: 0.9725 - val_loss: 0.0618 - val_acc: 0.9807\n",
      "Epoch 49/1536\n",
      "Epoch 00048: val_loss did not improve\n",
      "0s - loss: 0.0305 - acc: 0.9905 - val_loss: 0.0596 - val_acc: 0.9809\n",
      "Epoch 50/1536\n",
      "Epoch 00049: val_loss did not improve\n",
      "0s - loss: 0.0292 - acc: 0.9902 - val_loss: 0.0589 - val_acc: 0.9812\n",
      "Epoch 51/1536\n",
      "Epoch 00050: val_loss did not improve\n",
      "0s - loss: 0.0282 - acc: 0.9903 - val_loss: 0.0998 - val_acc: 0.9699\n",
      "Epoch 52/1536\n",
      "Epoch 00051: val_loss did not improve\n",
      "0s - loss: 0.4077 - acc: 0.9164 - val_loss: 1.7263 - val_acc: 0.7740\n",
      "Epoch 53/1536\n",
      "Epoch 00052: val_loss did not improve\n",
      "0s - loss: 2.0529 - acc: 0.7467 - val_loss: 0.0819 - val_acc: 0.9744\n",
      "Epoch 54/1536\n",
      "Epoch 00053: val_loss did not improve\n",
      "0s - loss: 0.0519 - acc: 0.9826 - val_loss: 0.0705 - val_acc: 0.9773\n",
      "Epoch 55/1536\n",
      "Epoch 00054: val_loss did not improve\n",
      "0s - loss: 0.0426 - acc: 0.9862 - val_loss: 0.0670 - val_acc: 0.9787\n",
      "Epoch 56/1536\n",
      "Epoch 00055: val_loss did not improve\n",
      "0s - loss: 0.0384 - acc: 0.9869 - val_loss: 0.0655 - val_acc: 0.9792\n",
      "Epoch 57/1536\n",
      "Epoch 00056: val_loss did not improve\n",
      "0s - loss: 0.0343 - acc: 0.9883 - val_loss: 0.0712 - val_acc: 0.9782\n",
      "Epoch 58/1536\n",
      "Epoch 00057: val_loss did not improve\n",
      "0s - loss: 0.0442 - acc: 0.9846 - val_loss: 0.1012 - val_acc: 0.9694\n",
      "Epoch 59/1536\n",
      "Epoch 00058: val_loss did not improve\n",
      "0s - loss: 0.0845 - acc: 0.9716 - val_loss: 0.1039 - val_acc: 0.9683\n",
      "Epoch 60/1536\n",
      "Epoch 00059: val_loss did not improve\n",
      "0s - loss: 0.5012 - acc: 0.8947 - val_loss: 1.9712 - val_acc: 0.6962\n",
      "Epoch 61/1536\n",
      "Epoch 00060: val_loss did not improve\n",
      "0s - loss: 1.3975 - acc: 0.8061 - val_loss: 0.0815 - val_acc: 0.9756\n",
      "Epoch 62/1536\n",
      "Epoch 00061: val_loss did not improve\n",
      "0s - loss: 0.0441 - acc: 0.9850 - val_loss: 0.0701 - val_acc: 0.9783\n",
      "Epoch 63/1536\n",
      "Epoch 00062: val_loss did not improve\n",
      "0s - loss: 0.0365 - acc: 0.9882 - val_loss: 0.0699 - val_acc: 0.9776\n",
      "Epoch 64/1536\n",
      "Epoch 00063: val_loss did not improve\n",
      "0s - loss: 0.0337 - acc: 0.9884 - val_loss: 0.0678 - val_acc: 0.9787\n",
      "Epoch 65/1536\n",
      "Epoch 00064: val_loss did not improve\n",
      "0s - loss: 0.0314 - acc: 0.9894 - val_loss: 0.0667 - val_acc: 0.9784\n",
      "Epoch 66/1536\n",
      "Epoch 00065: val_loss did not improve\n",
      "0s - loss: 0.0314 - acc: 0.9895 - val_loss: 0.1144 - val_acc: 0.9650\n",
      "Epoch 67/1536\n",
      "Epoch 00066: val_loss did not improve\n",
      "0s - loss: 0.4606 - acc: 0.9086 - val_loss: 1.5471 - val_acc: 0.7663\n",
      "Epoch 68/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00067: val_loss did not improve\n",
      "0s - loss: 0.9674 - acc: 0.8531 - val_loss: 0.0711 - val_acc: 0.9772\n",
      "Epoch 69/1536\n",
      "Epoch 00068: val_loss did not improve\n",
      "0s - loss: 0.0397 - acc: 0.9863 - val_loss: 0.0665 - val_acc: 0.9784\n",
      "Epoch 70/1536\n",
      "Epoch 00069: val_loss did not improve\n",
      "0s - loss: 0.0344 - acc: 0.9881 - val_loss: 0.0697 - val_acc: 0.9774\n",
      "Epoch 71/1536\n",
      "Epoch 00070: val_loss did not improve\n",
      "0s - loss: 0.0307 - acc: 0.9896 - val_loss: 0.0638 - val_acc: 0.9784\n",
      "Epoch 72/1536\n",
      "Epoch 00071: val_loss did not improve\n",
      "0s - loss: 0.0362 - acc: 0.9873 - val_loss: 0.2152 - val_acc: 0.9420\n",
      "Epoch 73/1536\n",
      "Epoch 00072: val_loss did not improve\n",
      "0s - loss: 0.1965 - acc: 0.9412 - val_loss: 0.0788 - val_acc: 0.9747\n",
      "Epoch 74/1536\n",
      "Epoch 00073: val_loss did not improve\n",
      "0s - loss: 0.0325 - acc: 0.9886 - val_loss: 0.0643 - val_acc: 0.9797\n",
      "Epoch 75/1536\n",
      "Epoch 00074: val_loss did not improve\n",
      "0s - loss: 0.0266 - acc: 0.9910 - val_loss: 0.1040 - val_acc: 0.9671\n",
      "Epoch 76/1536\n",
      "Epoch 00075: val_loss did not improve\n",
      "0s - loss: 0.9024 - acc: 0.8718 - val_loss: 2.6590 - val_acc: 0.7039\n",
      "Epoch 77/1536\n",
      "Epoch 00076: val_loss did not improve\n",
      "0s - loss: 1.4747 - acc: 0.8136 - val_loss: 0.0767 - val_acc: 0.9768\n",
      "Epoch 78/1536\n",
      "Epoch 00077: val_loss did not improve\n",
      "0s - loss: 0.0425 - acc: 0.9852 - val_loss: 0.0689 - val_acc: 0.9785\n",
      "Epoch 79/1536\n",
      "Epoch 00078: val_loss did not improve\n",
      "0s - loss: 0.0377 - acc: 0.9875 - val_loss: 0.0670 - val_acc: 0.9792\n",
      "Epoch 80/1536\n",
      "Epoch 00079: val_loss did not improve\n",
      "0s - loss: 0.0347 - acc: 0.9881 - val_loss: 0.0652 - val_acc: 0.9800\n",
      "Epoch 81/1536\n",
      "Epoch 00080: val_loss did not improve\n",
      "0s - loss: 0.0289 - acc: 0.9902 - val_loss: 0.0665 - val_acc: 0.9802\n",
      "Epoch 82/1536\n",
      "Epoch 00081: val_loss did not improve\n",
      "0s - loss: 0.0297 - acc: 0.9893 - val_loss: 0.0759 - val_acc: 0.9772\n",
      "Epoch 83/1536\n",
      "Epoch 00082: val_loss did not improve\n",
      "0s - loss: 0.0509 - acc: 0.9821 - val_loss: 0.5862 - val_acc: 0.8888\n",
      "Epoch 84/1536\n",
      "Epoch 00083: val_loss did not improve\n",
      "0s - loss: 0.9433 - acc: 0.8454 - val_loss: 0.8598 - val_acc: 0.8447\n",
      "Epoch 85/1536\n",
      "Epoch 00084: val_loss did not improve\n",
      "0s - loss: 0.2969 - acc: 0.9351 - val_loss: 0.0704 - val_acc: 0.9801\n",
      "Epoch 86/1536\n",
      "Epoch 00085: val_loss did not improve\n",
      "0s - loss: 0.0335 - acc: 0.9889 - val_loss: 0.0679 - val_acc: 0.9807\n",
      "Epoch 87/1536\n",
      "Epoch 00086: val_loss did not improve\n",
      "0s - loss: 0.0304 - acc: 0.9894 - val_loss: 0.0661 - val_acc: 0.9807\n",
      "Epoch 88/1536\n",
      "Epoch 00087: val_loss did not improve\n",
      "0s - loss: 0.0272 - acc: 0.9908 - val_loss: 0.0624 - val_acc: 0.9817\n",
      "Epoch 89/1536\n",
      "Epoch 00088: val_loss did not improve\n",
      "0s - loss: 0.0281 - acc: 0.9906 - val_loss: 0.0850 - val_acc: 0.9744\n",
      "Epoch 90/1536\n",
      "Epoch 00089: val_loss did not improve\n",
      "0s - loss: 0.3674 - acc: 0.9351 - val_loss: 1.3607 - val_acc: 0.7958\n",
      "Epoch 91/1536\n",
      "Epoch 00090: val_loss did not improve\n",
      "0s - loss: 1.2512 - acc: 0.8049 - val_loss: 0.0832 - val_acc: 0.9758\n",
      "Epoch 92/1536\n",
      "Epoch 00091: val_loss did not improve\n",
      "0s - loss: 0.0385 - acc: 0.9863 - val_loss: 0.0691 - val_acc: 0.9778\n",
      "Epoch 93/1536\n",
      "Epoch 00092: val_loss did not improve\n",
      "0s - loss: 0.0315 - acc: 0.9891 - val_loss: 0.0684 - val_acc: 0.9787\n",
      "Epoch 94/1536\n",
      "Epoch 00093: val_loss did not improve\n",
      "0s - loss: 0.0290 - acc: 0.9898 - val_loss: 0.0662 - val_acc: 0.9786\n",
      "Epoch 95/1536\n",
      "Epoch 00094: val_loss did not improve\n",
      "0s - loss: 0.0279 - acc: 0.9901 - val_loss: 0.0636 - val_acc: 0.9800\n",
      "Epoch 96/1536\n",
      "Epoch 00095: val_loss did not improve\n",
      "0s - loss: 0.0267 - acc: 0.9904 - val_loss: 0.0713 - val_acc: 0.9786\n",
      "Epoch 97/1536\n",
      "Epoch 00096: val_loss did not improve\n",
      "0s - loss: 0.0852 - acc: 0.9736 - val_loss: 0.8308 - val_acc: 0.8556\n",
      "Epoch 98/1536\n",
      "Epoch 00097: val_loss did not improve\n",
      "0s - loss: 1.0873 - acc: 0.8289 - val_loss: 0.4355 - val_acc: 0.9046\n",
      "Epoch 99/1536\n",
      "Epoch 00098: val_loss did not improve\n",
      "0s - loss: 0.1300 - acc: 0.9671 - val_loss: 0.0703 - val_acc: 0.9784\n",
      "Epoch 100/1536\n",
      "Epoch 00099: val_loss did not improve\n",
      "0s - loss: 0.0331 - acc: 0.9891 - val_loss: 0.0723 - val_acc: 0.9774\n",
      "Epoch 101/1536\n",
      "Epoch 00100: val_loss did not improve\n",
      "0s - loss: 0.0301 - acc: 0.9896 - val_loss: 0.0641 - val_acc: 0.9798\n",
      "Epoch 102/1536\n",
      "Epoch 00101: val_loss did not improve\n",
      "0s - loss: 0.0288 - acc: 0.9900 - val_loss: 0.0621 - val_acc: 0.9801\n",
      "Epoch 103/1536\n",
      "Epoch 00102: val_loss did not improve\n",
      "0s - loss: 0.0451 - acc: 0.9843 - val_loss: 0.5636 - val_acc: 0.8921\n",
      "Epoch 104/1536\n",
      "Epoch 00103: val_loss did not improve\n",
      "0s - loss: 0.9950 - acc: 0.8332 - val_loss: 0.1510 - val_acc: 0.9578\n",
      "Epoch 105/1536\n",
      "Epoch 00104: val_loss did not improve\n",
      "0s - loss: 0.0561 - acc: 0.9814 - val_loss: 0.0669 - val_acc: 0.9796\n",
      "Epoch 106/1536\n",
      "Epoch 00105: val_loss did not improve\n",
      "0s - loss: 0.0300 - acc: 0.9897 - val_loss: 0.0613 - val_acc: 0.9812\n",
      "Epoch 107/1536\n",
      "Epoch 00106: val_loss did not improve\n",
      "0s - loss: 0.0299 - acc: 0.9897 - val_loss: 0.0720 - val_acc: 0.9781\n",
      "Epoch 108/1536\n",
      "Epoch 00107: val_loss did not improve\n",
      "0s - loss: 0.0475 - acc: 0.9834 - val_loss: 0.2569 - val_acc: 0.9267\n",
      "Epoch 109/1536\n",
      "Epoch 00108: val_loss did not improve\n",
      "0s - loss: 0.3048 - acc: 0.9181 - val_loss: 1.1424 - val_acc: 0.8442\n",
      "Epoch 110/1536\n",
      "Epoch 00109: val_loss did not improve\n",
      "0s - loss: 1.1556 - acc: 0.8241 - val_loss: 0.0891 - val_acc: 0.9729\n",
      "Epoch 111/1536\n",
      "Epoch 00110: val_loss did not improve\n",
      "0s - loss: 0.0406 - acc: 0.9858 - val_loss: 0.0735 - val_acc: 0.9772\n",
      "Epoch 112/1536\n",
      "Epoch 00111: val_loss did not improve\n",
      "0s - loss: 0.0329 - acc: 0.9888 - val_loss: 0.0642 - val_acc: 0.9800\n",
      "Epoch 113/1536\n",
      "Epoch 00112: val_loss did not improve\n",
      "0s - loss: 0.0292 - acc: 0.9899 - val_loss: 0.0676 - val_acc: 0.9799\n",
      "Epoch 114/1536\n",
      "Epoch 00113: val_loss did not improve\n",
      "0s - loss: 0.0285 - acc: 0.9902 - val_loss: 0.0648 - val_acc: 0.9801\n",
      "Epoch 115/1536\n",
      "Epoch 00114: val_loss did not improve\n",
      "0s - loss: 0.0324 - acc: 0.9889 - val_loss: 0.1084 - val_acc: 0.9647\n",
      "Epoch 116/1536\n",
      "Epoch 00115: val_loss did not improve\n",
      "0s - loss: 0.4112 - acc: 0.9171 - val_loss: 0.3774 - val_acc: 0.9060\n",
      "Epoch 117/1536\n",
      "Epoch 00116: val_loss did not improve\n",
      "0s - loss: 0.2516 - acc: 0.9311 - val_loss: 0.0637 - val_acc: 0.9793\n",
      "Epoch 118/1536\n",
      "Epoch 00117: val_loss did not improve\n",
      "0s - loss: 0.0278 - acc: 0.9903 - val_loss: 0.0583 - val_acc: 0.9818\n",
      "Epoch 119/1536\n",
      "Epoch 00118: val_loss did not improve\n",
      "0s - loss: 0.0255 - acc: 0.9910 - val_loss: 0.0570 - val_acc: 0.9820\n",
      "Epoch 120/1536\n",
      "Epoch 00119: val_loss did not improve\n",
      "0s - loss: 0.0304 - acc: 0.9899 - val_loss: 0.3539 - val_acc: 0.9255\n",
      "Epoch 121/1536\n",
      "Epoch 00120: val_loss did not improve\n",
      "0s - loss: 1.2422 - acc: 0.8217 - val_loss: 1.6152 - val_acc: 0.7574\n",
      "Epoch 122/1536\n",
      "Epoch 00121: val_loss did not improve\n",
      "0s - loss: 0.7545 - acc: 0.8767 - val_loss: 0.0778 - val_acc: 0.9772\n",
      "Epoch 123/1536\n",
      "Epoch 00122: val_loss did not improve\n",
      "0s - loss: 0.0397 - acc: 0.9862 - val_loss: 0.0703 - val_acc: 0.9792\n",
      "Epoch 124/1536\n",
      "Epoch 00123: val_loss did not improve\n",
      "0s - loss: 0.0329 - acc: 0.9885 - val_loss: 0.0680 - val_acc: 0.9801\n",
      "Epoch 125/1536\n",
      "Epoch 00124: val_loss did not improve\n",
      "0s - loss: 0.0302 - acc: 0.9899 - val_loss: 0.0667 - val_acc: 0.9798\n",
      "Epoch 126/1536\n",
      "Epoch 00125: val_loss did not improve\n",
      "0s - loss: 0.0291 - acc: 0.9902 - val_loss: 0.0655 - val_acc: 0.9799\n",
      "Epoch 127/1536\n",
      "Epoch 00126: val_loss did not improve\n",
      "0s - loss: 0.0261 - acc: 0.9915 - val_loss: 0.0820 - val_acc: 0.9755\n",
      "Epoch 128/1536\n",
      "Epoch 00127: val_loss did not improve\n",
      "0s - loss: 0.1451 - acc: 0.9611 - val_loss: 0.4728 - val_acc: 0.8904\n",
      "Epoch 129/1536\n",
      "Epoch 00128: val_loss did not improve\n",
      "0s - loss: 1.0517 - acc: 0.8320 - val_loss: 0.6268 - val_acc: 0.8653\n",
      "Epoch 130/1536\n",
      "Epoch 00129: val_loss did not improve\n",
      "0s - loss: 0.2555 - acc: 0.9423 - val_loss: 0.0684 - val_acc: 0.9793\n",
      "Epoch 131/1536\n",
      "Epoch 00130: val_loss did not improve\n",
      "0s - loss: 0.0307 - acc: 0.9896 - val_loss: 0.0654 - val_acc: 0.9794\n",
      "Epoch 132/1536\n",
      "Epoch 00131: val_loss did not improve\n",
      "0s - loss: 0.0274 - acc: 0.9908 - val_loss: 0.0601 - val_acc: 0.9808\n",
      "Epoch 133/1536\n",
      "Epoch 00132: val_loss did not improve\n",
      "0s - loss: 0.0253 - acc: 0.9911 - val_loss: 0.0609 - val_acc: 0.9802\n",
      "Epoch 134/1536\n",
      "Epoch 00133: val_loss did not improve\n",
      "0s - loss: 0.0322 - acc: 0.9891 - val_loss: 0.1398 - val_acc: 0.9596\n",
      "Epoch 135/1536\n",
      "Epoch 00134: val_loss did not improve\n",
      "0s - loss: 0.4387 - acc: 0.9048 - val_loss: 0.5116 - val_acc: 0.9008\n",
      "Epoch 136/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00135: val_loss did not improve\n",
      "0s - loss: 0.1914 - acc: 0.9507 - val_loss: 0.0631 - val_acc: 0.9812\n",
      "Epoch 137/1536\n",
      "Epoch 00136: val_loss did not improve\n",
      "0s - loss: 0.0254 - acc: 0.9914 - val_loss: 0.0658 - val_acc: 0.9802\n",
      "Epoch 138/1536\n",
      "Epoch 00137: val_loss did not improve\n",
      "0s - loss: 0.0269 - acc: 0.9912 - val_loss: 0.0734 - val_acc: 0.9776\n",
      "Epoch 139/1536\n",
      "Epoch 00138: val_loss did not improve\n",
      "0s - loss: 0.0325 - acc: 0.9890 - val_loss: 0.0725 - val_acc: 0.9779\n",
      "Epoch 140/1536\n",
      "Epoch 00139: val_loss did not improve\n",
      "0s - loss: 0.1132 - acc: 0.9682 - val_loss: 0.7311 - val_acc: 0.8562\n",
      "Epoch 141/1536\n",
      "Epoch 00140: val_loss did not improve\n",
      "0s - loss: 1.3693 - acc: 0.7893 - val_loss: 0.1095 - val_acc: 0.9679\n",
      "Epoch 142/1536\n",
      "Epoch 00141: val_loss did not improve\n",
      "0s - loss: 0.0486 - acc: 0.9841 - val_loss: 0.0652 - val_acc: 0.9797\n",
      "Epoch 143/1536\n",
      "Epoch 00142: val_loss did not improve\n",
      "0s - loss: 0.0320 - acc: 0.9890 - val_loss: 0.0631 - val_acc: 0.9808\n",
      "Epoch 144/1536\n",
      "Epoch 00143: val_loss did not improve\n",
      "0s - loss: 0.0288 - acc: 0.9909 - val_loss: 0.0616 - val_acc: 0.9802\n",
      "Epoch 145/1536\n",
      "Epoch 00144: val_loss did not improve\n",
      "0s - loss: 0.0257 - acc: 0.9915 - val_loss: 0.0615 - val_acc: 0.9812\n",
      "Epoch 146/1536\n",
      "Epoch 00145: val_loss did not improve\n",
      "0s - loss: 0.0255 - acc: 0.9918 - val_loss: 0.0683 - val_acc: 0.9795\n",
      "Epoch 147/1536\n",
      "Epoch 00146: val_loss did not improve\n",
      "0s - loss: 0.1401 - acc: 0.9625 - val_loss: 0.5607 - val_acc: 0.8756\n",
      "Epoch 148/1536\n",
      "Epoch 00147: val_loss did not improve\n",
      "0s - loss: 1.1141 - acc: 0.8151 - val_loss: 0.7841 - val_acc: 0.8493\n",
      "Epoch 149/1536\n",
      "Epoch 00148: val_loss did not improve\n",
      "0s - loss: 0.2751 - acc: 0.9382 - val_loss: 0.0644 - val_acc: 0.9795\n",
      "Epoch 150/1536\n",
      "Epoch 00149: val_loss did not improve\n",
      "0s - loss: 0.0289 - acc: 0.9900 - val_loss: 0.0602 - val_acc: 0.9811\n",
      "Epoch 151/1536\n",
      "Epoch 00150: val_loss did not improve\n",
      "0s - loss: 0.0279 - acc: 0.9903 - val_loss: 0.0586 - val_acc: 0.9817\n",
      "Epoch 152/1536\n",
      "Epoch 00151: val_loss did not improve\n",
      "0s - loss: 0.0249 - acc: 0.9920 - val_loss: 0.0630 - val_acc: 0.9798\n",
      "Epoch 153/1536\n",
      "Epoch 00152: val_loss did not improve\n",
      "0s - loss: 0.0261 - acc: 0.9906 - val_loss: 0.0623 - val_acc: 0.9793\n",
      "Epoch 154/1536\n",
      "Epoch 00153: val_loss did not improve\n",
      "0s - loss: 0.0871 - acc: 0.9744 - val_loss: 0.4310 - val_acc: 0.9188\n",
      "Epoch 155/1536\n",
      "Epoch 00154: val_loss did not improve\n",
      "0s - loss: 0.4955 - acc: 0.8968 - val_loss: 0.3953 - val_acc: 0.9180\n",
      "Epoch 156/1536\n",
      "Epoch 00155: val_loss did not improve\n",
      "0s - loss: 0.1411 - acc: 0.9633 - val_loss: 0.0615 - val_acc: 0.9799\n",
      "Epoch 157/1536\n",
      "Epoch 00156: val_loss did not improve\n",
      "0s - loss: 0.0254 - acc: 0.9916 - val_loss: 0.0666 - val_acc: 0.9775\n",
      "Epoch 158/1536\n",
      "Epoch 00157: val_loss did not improve\n",
      "0s - loss: 0.0240 - acc: 0.9918 - val_loss: 0.0651 - val_acc: 0.9790\n",
      "Epoch 159/1536\n",
      "Epoch 00158: val_loss did not improve\n",
      "0s - loss: 0.0232 - acc: 0.9925 - val_loss: 0.0794 - val_acc: 0.9757\n",
      "Epoch 160/1536\n",
      "Epoch 00159: val_loss did not improve\n",
      "0s - loss: 0.5383 - acc: 0.8906 - val_loss: 0.7130 - val_acc: 0.8528\n",
      "Epoch 161/1536\n",
      "Epoch 00160: val_loss did not improve\n",
      "0s - loss: 0.4154 - acc: 0.9117 - val_loss: 0.0773 - val_acc: 0.9770\n",
      "Epoch 162/1536\n",
      "Epoch 00161: val_loss did not improve\n",
      "0s - loss: 0.0325 - acc: 0.9889 - val_loss: 0.0705 - val_acc: 0.9786\n",
      "Epoch 163/1536\n",
      "Epoch 00162: val_loss did not improve\n",
      "0s - loss: 0.0291 - acc: 0.9903 - val_loss: 0.0685 - val_acc: 0.9787\n",
      "Epoch 164/1536\n",
      "Epoch 00163: val_loss did not improve\n",
      "0s - loss: 0.0273 - acc: 0.9911 - val_loss: 0.0710 - val_acc: 0.9775\n",
      "Epoch 165/1536\n",
      "Epoch 00164: val_loss did not improve\n",
      "0s - loss: 0.0474 - acc: 0.9842 - val_loss: 1.1721 - val_acc: 0.8454\n",
      "Epoch 166/1536\n",
      "Epoch 00165: val_loss did not improve\n",
      "0s - loss: 2.2910 - acc: 0.7194 - val_loss: 0.2296 - val_acc: 0.9416\n",
      "Epoch 167/1536\n",
      "Epoch 00166: val_loss did not improve\n",
      "0s - loss: 0.0899 - acc: 0.9718 - val_loss: 0.0744 - val_acc: 0.9783\n",
      "Epoch 168/1536\n",
      "Epoch 00167: val_loss did not improve\n",
      "0s - loss: 0.0417 - acc: 0.9856 - val_loss: 0.0684 - val_acc: 0.9794\n",
      "Epoch 169/1536\n",
      "Epoch 00168: val_loss did not improve\n",
      "0s - loss: 0.0349 - acc: 0.9883 - val_loss: 0.0637 - val_acc: 0.9806\n",
      "Epoch 170/1536\n",
      "Epoch 00169: val_loss did not improve\n",
      "0s - loss: 0.0309 - acc: 0.9892 - val_loss: 0.0628 - val_acc: 0.9810\n",
      "Epoch 171/1536\n",
      "Epoch 00170: val_loss did not improve\n",
      "0s - loss: 0.0286 - acc: 0.9902 - val_loss: 0.0596 - val_acc: 0.9820\n",
      "Epoch 172/1536\n",
      "Epoch 00171: val_loss did not improve\n",
      "0s - loss: 0.0289 - acc: 0.9898 - val_loss: 0.0696 - val_acc: 0.9781\n",
      "Epoch 173/1536\n",
      "Epoch 00172: val_loss did not improve\n",
      "0s - loss: 0.2270 - acc: 0.9549 - val_loss: 0.9698 - val_acc: 0.8205\n",
      "Epoch 174/1536\n",
      "Epoch 00173: val_loss did not improve\n",
      "0s - loss: 0.8387 - acc: 0.8518 - val_loss: 0.0735 - val_acc: 0.9769\n",
      "Epoch 175/1536\n",
      "Epoch 00174: val_loss did not improve\n",
      "0s - loss: 0.0310 - acc: 0.9893 - val_loss: 0.0658 - val_acc: 0.9806\n",
      "Epoch 176/1536\n",
      "Epoch 00175: val_loss did not improve\n",
      "0s - loss: 0.0309 - acc: 0.9896 - val_loss: 0.0681 - val_acc: 0.9791\n",
      "Epoch 177/1536\n",
      "Epoch 00176: val_loss did not improve\n",
      "0s - loss: 0.0275 - acc: 0.9905 - val_loss: 0.0608 - val_acc: 0.9813\n",
      "Epoch 178/1536\n",
      "Epoch 00177: val_loss did not improve\n",
      "0s - loss: 0.1099 - acc: 0.9680 - val_loss: 0.8655 - val_acc: 0.8517\n",
      "Epoch 179/1536\n",
      "Epoch 00178: val_loss did not improve\n",
      "0s - loss: 0.5374 - acc: 0.8913 - val_loss: 0.2374 - val_acc: 0.9340\n",
      "Epoch 180/1536\n",
      "Epoch 00179: val_loss improved from 0.05668 to 0.05442, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0945 - acc: 0.9733 - val_loss: 0.0544 - val_acc: 0.9820\n",
      "Epoch 181/1536\n",
      "Epoch 00180: val_loss did not improve\n",
      "0s - loss: 0.0243 - acc: 0.9921 - val_loss: 0.0587 - val_acc: 0.9808\n",
      "Epoch 182/1536\n",
      "Epoch 00181: val_loss did not improve\n",
      "0s - loss: 0.0239 - acc: 0.9920 - val_loss: 0.0586 - val_acc: 0.9801\n",
      "Epoch 183/1536\n",
      "Epoch 00182: val_loss did not improve\n",
      "0s - loss: 0.0945 - acc: 0.9713 - val_loss: 0.3215 - val_acc: 0.9256\n",
      "Epoch 184/1536\n",
      "Epoch 00183: val_loss did not improve\n",
      "0s - loss: 0.3284 - acc: 0.9166 - val_loss: 0.3734 - val_acc: 0.9074\n",
      "Epoch 185/1536\n",
      "Epoch 00184: val_loss did not improve\n",
      "0s - loss: 0.1604 - acc: 0.9570 - val_loss: 0.0579 - val_acc: 0.9821\n",
      "Epoch 186/1536\n",
      "Epoch 00185: val_loss did not improve\n",
      "0s - loss: 0.0230 - acc: 0.9927 - val_loss: 0.0571 - val_acc: 0.9819\n",
      "Epoch 187/1536\n",
      "Epoch 00186: val_loss did not improve\n",
      "0s - loss: 0.0236 - acc: 0.9924 - val_loss: 0.0685 - val_acc: 0.9795\n",
      "Epoch 188/1536\n",
      "Epoch 00187: val_loss did not improve\n",
      "0s - loss: 0.2043 - acc: 0.9531 - val_loss: 0.4463 - val_acc: 0.9003\n",
      "Epoch 189/1536\n",
      "Epoch 00188: val_loss did not improve\n",
      "0s - loss: 0.5737 - acc: 0.8732 - val_loss: 0.6234 - val_acc: 0.8868\n",
      "Epoch 190/1536\n",
      "Epoch 00189: val_loss did not improve\n",
      "0s - loss: 0.2552 - acc: 0.9423 - val_loss: 0.0614 - val_acc: 0.9810\n",
      "Epoch 191/1536\n",
      "Epoch 00190: val_loss did not improve\n",
      "0s - loss: 0.0280 - acc: 0.9911 - val_loss: 0.0599 - val_acc: 0.9817\n",
      "Epoch 192/1536\n",
      "Epoch 00191: val_loss did not improve\n",
      "0s - loss: 0.0268 - acc: 0.9914 - val_loss: 0.0583 - val_acc: 0.9815\n",
      "Epoch 193/1536\n",
      "Epoch 00192: val_loss did not improve\n",
      "0s - loss: 0.0273 - acc: 0.9908 - val_loss: 0.0630 - val_acc: 0.9807\n",
      "Epoch 194/1536\n",
      "Epoch 00193: val_loss did not improve\n",
      "0s - loss: 0.0354 - acc: 0.9882 - val_loss: 0.2719 - val_acc: 0.9350\n",
      "Epoch 195/1536\n",
      "Epoch 00194: val_loss did not improve\n",
      "0s - loss: 0.6116 - acc: 0.8783 - val_loss: 0.5439 - val_acc: 0.8775\n",
      "Epoch 196/1536\n",
      "Epoch 00195: val_loss did not improve\n",
      "0s - loss: 0.2632 - acc: 0.9432 - val_loss: 0.0613 - val_acc: 0.9802\n",
      "Epoch 197/1536\n",
      "Epoch 00196: val_loss did not improve\n",
      "0s - loss: 0.0289 - acc: 0.9911 - val_loss: 0.0582 - val_acc: 0.9813\n",
      "Epoch 198/1536\n",
      "Epoch 00197: val_loss did not improve\n",
      "0s - loss: 0.0240 - acc: 0.9925 - val_loss: 0.0569 - val_acc: 0.9815\n",
      "Epoch 199/1536\n",
      "Epoch 00198: val_loss did not improve\n",
      "0s - loss: 0.0243 - acc: 0.9925 - val_loss: 0.0748 - val_acc: 0.9759\n",
      "Epoch 200/1536\n",
      "Epoch 00199: val_loss did not improve\n",
      "0s - loss: 0.2989 - acc: 0.9351 - val_loss: 0.9928 - val_acc: 0.8305\n",
      "Epoch 201/1536\n",
      "Epoch 00200: val_loss did not improve\n",
      "0s - loss: 0.8089 - acc: 0.8623 - val_loss: 0.0705 - val_acc: 0.9788\n",
      "Epoch 202/1536\n",
      "Epoch 00201: val_loss did not improve\n",
      "0s - loss: 0.0317 - acc: 0.9896 - val_loss: 0.0677 - val_acc: 0.9795\n",
      "Epoch 203/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00202: val_loss did not improve\n",
      "0s - loss: 0.0285 - acc: 0.9907 - val_loss: 0.0641 - val_acc: 0.9806\n",
      "Epoch 204/1536\n",
      "Epoch 00203: val_loss did not improve\n",
      "0s - loss: 0.0252 - acc: 0.9917 - val_loss: 0.0600 - val_acc: 0.9820\n",
      "Epoch 205/1536\n",
      "Epoch 00204: val_loss did not improve\n",
      "0s - loss: 0.0246 - acc: 0.9918 - val_loss: 0.0880 - val_acc: 0.9730\n",
      "Epoch 206/1536\n",
      "Epoch 00205: val_loss did not improve\n",
      "0s - loss: 0.1903 - acc: 0.9544 - val_loss: 0.4475 - val_acc: 0.8975\n",
      "Epoch 207/1536\n",
      "Epoch 00206: val_loss did not improve\n",
      "0s - loss: 0.5389 - acc: 0.8788 - val_loss: 0.1277 - val_acc: 0.9626\n",
      "Epoch 208/1536\n",
      "Epoch 00207: val_loss did not improve\n",
      "0s - loss: 0.0410 - acc: 0.9864 - val_loss: 0.0593 - val_acc: 0.9809\n",
      "Epoch 209/1536\n",
      "Epoch 00208: val_loss did not improve\n",
      "0s - loss: 0.0267 - acc: 0.9914 - val_loss: 0.0608 - val_acc: 0.9808\n",
      "Epoch 210/1536\n",
      "Epoch 00209: val_loss did not improve\n",
      "0s - loss: 0.0245 - acc: 0.9922 - val_loss: 0.0643 - val_acc: 0.9788\n",
      "Epoch 211/1536\n",
      "Epoch 00210: val_loss did not improve\n",
      "0s - loss: 0.0271 - acc: 0.9906 - val_loss: 0.0650 - val_acc: 0.9795\n",
      "Epoch 212/1536\n",
      "Epoch 00211: val_loss did not improve\n",
      "0s - loss: 0.0435 - acc: 0.9854 - val_loss: 0.8133 - val_acc: 0.9059\n",
      "Epoch 213/1536\n",
      "Epoch 00212: val_loss did not improve\n",
      "0s - loss: 1.1820 - acc: 0.8149 - val_loss: 0.6305 - val_acc: 0.8662\n",
      "Epoch 214/1536\n",
      "Epoch 00213: val_loss did not improve\n",
      "0s - loss: 0.2850 - acc: 0.9349 - val_loss: 0.0639 - val_acc: 0.9804\n",
      "Epoch 215/1536\n",
      "Epoch 00214: val_loss did not improve\n",
      "0s - loss: 0.0283 - acc: 0.9904 - val_loss: 0.0635 - val_acc: 0.9797\n",
      "Epoch 216/1536\n",
      "Epoch 00215: val_loss did not improve\n",
      "0s - loss: 0.0260 - acc: 0.9911 - val_loss: 0.0604 - val_acc: 0.9817\n",
      "Epoch 217/1536\n",
      "Epoch 00216: val_loss did not improve\n",
      "0s - loss: 0.0231 - acc: 0.9922 - val_loss: 0.0639 - val_acc: 0.9792\n",
      "Epoch 218/1536\n",
      "Epoch 00217: val_loss did not improve\n",
      "0s - loss: 0.0232 - acc: 0.9922 - val_loss: 0.0579 - val_acc: 0.9818\n",
      "Epoch 219/1536\n",
      "Epoch 00218: val_loss did not improve\n",
      "0s - loss: 0.0247 - acc: 0.9919 - val_loss: 0.0720 - val_acc: 0.9776\n",
      "Epoch 220/1536\n",
      "Epoch 00219: val_loss did not improve\n",
      "0s - loss: 0.3013 - acc: 0.9450 - val_loss: 0.7530 - val_acc: 0.8725\n",
      "Epoch 221/1536\n",
      "Epoch 00220: val_loss did not improve\n",
      "0s - loss: 0.3978 - acc: 0.9158 - val_loss: 0.0625 - val_acc: 0.9808\n",
      "Epoch 222/1536\n",
      "Epoch 00221: val_loss did not improve\n",
      "0s - loss: 0.0266 - acc: 0.9916 - val_loss: 0.0606 - val_acc: 0.9822\n",
      "Epoch 223/1536\n",
      "Epoch 00222: val_loss did not improve\n",
      "0s - loss: 0.0257 - acc: 0.9911 - val_loss: 0.0572 - val_acc: 0.9824\n",
      "Epoch 224/1536\n",
      "Epoch 00223: val_loss did not improve\n",
      "0s - loss: 0.0229 - acc: 0.9921 - val_loss: 0.0776 - val_acc: 0.9765\n",
      "Epoch 225/1536\n",
      "Epoch 00224: val_loss did not improve\n",
      "0s - loss: 0.3823 - acc: 0.9332 - val_loss: 1.1606 - val_acc: 0.8216\n",
      "Epoch 226/1536\n",
      "Epoch 00225: val_loss did not improve\n",
      "0s - loss: 1.6211 - acc: 0.7860 - val_loss: 0.0857 - val_acc: 0.9752\n",
      "Epoch 227/1536\n",
      "Epoch 00226: val_loss did not improve\n",
      "0s - loss: 0.0437 - acc: 0.9853 - val_loss: 0.0656 - val_acc: 0.9808\n",
      "Epoch 228/1536\n",
      "Epoch 00227: val_loss did not improve\n",
      "0s - loss: 0.0315 - acc: 0.9891 - val_loss: 0.0621 - val_acc: 0.9803\n",
      "Epoch 229/1536\n",
      "Epoch 00228: val_loss did not improve\n",
      "0s - loss: 0.0283 - acc: 0.9899 - val_loss: 0.0589 - val_acc: 0.9827\n",
      "Epoch 230/1536\n",
      "Epoch 00229: val_loss did not improve\n",
      "0s - loss: 0.0262 - acc: 0.9913 - val_loss: 0.0592 - val_acc: 0.9818\n",
      "Epoch 231/1536\n",
      "Epoch 00230: val_loss did not improve\n",
      "0s - loss: 0.0264 - acc: 0.9910 - val_loss: 0.0691 - val_acc: 0.9780\n",
      "Epoch 232/1536\n",
      "Epoch 00231: val_loss did not improve\n",
      "0s - loss: 0.1022 - acc: 0.9750 - val_loss: 0.2325 - val_acc: 0.9401\n",
      "Epoch 233/1536\n",
      "Epoch 00232: val_loss did not improve\n",
      "0s - loss: 0.4870 - acc: 0.9046 - val_loss: 0.3029 - val_acc: 0.9235\n",
      "Epoch 234/1536\n",
      "Epoch 00233: val_loss did not improve\n",
      "0s - loss: 0.1085 - acc: 0.9682 - val_loss: 0.0634 - val_acc: 0.9797\n",
      "Epoch 235/1536\n",
      "Epoch 00234: val_loss did not improve\n",
      "0s - loss: 0.0234 - acc: 0.9927 - val_loss: 0.0617 - val_acc: 0.9802\n",
      "Epoch 236/1536\n",
      "Epoch 00235: val_loss did not improve\n",
      "0s - loss: 0.0224 - acc: 0.9925 - val_loss: 0.0572 - val_acc: 0.9813\n",
      "Epoch 237/1536\n",
      "Epoch 00236: val_loss did not improve\n",
      "0s - loss: 0.0219 - acc: 0.9927 - val_loss: 0.0643 - val_acc: 0.9804\n",
      "Epoch 238/1536\n",
      "Epoch 00237: val_loss did not improve\n",
      "0s - loss: 0.1698 - acc: 0.9601 - val_loss: 0.5559 - val_acc: 0.8860\n",
      "Epoch 239/1536\n",
      "Epoch 00238: val_loss did not improve\n",
      "0s - loss: 1.0141 - acc: 0.8239 - val_loss: 0.1220 - val_acc: 0.9644\n",
      "Epoch 240/1536\n",
      "Epoch 00239: val_loss did not improve\n",
      "0s - loss: 0.0438 - acc: 0.9844 - val_loss: 0.0636 - val_acc: 0.9806\n",
      "Epoch 241/1536\n",
      "Epoch 00240: val_loss did not improve\n",
      "0s - loss: 0.0293 - acc: 0.9899 - val_loss: 0.0627 - val_acc: 0.9812\n",
      "Epoch 242/1536\n",
      "Epoch 00241: val_loss did not improve\n",
      "0s - loss: 0.0249 - acc: 0.9917 - val_loss: 0.0627 - val_acc: 0.9808\n",
      "Epoch 243/1536\n",
      "Epoch 00242: val_loss did not improve\n",
      "0s - loss: 0.0242 - acc: 0.9914 - val_loss: 0.0716 - val_acc: 0.9786\n",
      "Epoch 244/1536\n",
      "Epoch 00243: val_loss did not improve\n",
      "0s - loss: 0.0297 - acc: 0.9894 - val_loss: 0.1968 - val_acc: 0.9415\n",
      "Epoch 245/1536\n",
      "Epoch 00244: val_loss did not improve\n",
      "0s - loss: 0.5638 - acc: 0.8855 - val_loss: 0.4649 - val_acc: 0.8834\n",
      "Epoch 246/1536\n",
      "Epoch 00245: val_loss did not improve\n",
      "0s - loss: 0.1964 - acc: 0.9483 - val_loss: 0.0572 - val_acc: 0.9815\n",
      "Epoch 247/1536\n",
      "Epoch 00246: val_loss did not improve\n",
      "0s - loss: 0.0248 - acc: 0.9915 - val_loss: 0.0554 - val_acc: 0.9821\n",
      "Epoch 248/1536\n",
      "Epoch 00247: val_loss did not improve\n",
      "0s - loss: 0.0222 - acc: 0.9927 - val_loss: 0.0580 - val_acc: 0.9822\n",
      "Epoch 249/1536\n",
      "Epoch 00248: val_loss did not improve\n",
      "0s - loss: 0.0237 - acc: 0.9918 - val_loss: 0.0779 - val_acc: 0.9750\n",
      "Epoch 250/1536\n",
      "Epoch 00249: val_loss did not improve\n",
      "0s - loss: 0.2341 - acc: 0.9428 - val_loss: 0.7132 - val_acc: 0.8545\n",
      "Epoch 251/1536\n",
      "Epoch 00250: val_loss did not improve\n",
      "0s - loss: 0.3690 - acc: 0.9156 - val_loss: 0.0641 - val_acc: 0.9791\n",
      "Epoch 252/1536\n",
      "Epoch 00251: val_loss did not improve\n",
      "0s - loss: 0.0259 - acc: 0.9916 - val_loss: 0.0635 - val_acc: 0.9798\n",
      "Epoch 253/1536\n",
      "Epoch 00252: val_loss did not improve\n",
      "0s - loss: 0.0234 - acc: 0.9922 - val_loss: 0.0600 - val_acc: 0.9809\n",
      "Epoch 254/1536\n",
      "Epoch 00253: val_loss did not improve\n",
      "0s - loss: 0.0238 - acc: 0.9923 - val_loss: 0.1065 - val_acc: 0.9677\n",
      "Epoch 255/1536\n",
      "Epoch 00254: val_loss did not improve\n",
      "0s - loss: 0.7103 - acc: 0.8844 - val_loss: 2.1216 - val_acc: 0.7286\n",
      "Epoch 256/1536\n",
      "Epoch 00255: val_loss did not improve\n",
      "0s - loss: 1.4035 - acc: 0.8248 - val_loss: 0.0786 - val_acc: 0.9762\n",
      "Epoch 257/1536\n",
      "Epoch 00256: val_loss did not improve\n",
      "0s - loss: 0.0379 - acc: 0.9866 - val_loss: 0.0698 - val_acc: 0.9791\n",
      "Epoch 258/1536\n",
      "Epoch 00257: val_loss did not improve\n",
      "0s - loss: 0.0307 - acc: 0.9898 - val_loss: 0.0693 - val_acc: 0.9790\n",
      "Epoch 259/1536\n",
      "Epoch 00258: val_loss did not improve\n",
      "0s - loss: 0.0301 - acc: 0.9894 - val_loss: 0.0696 - val_acc: 0.9798\n",
      "Epoch 260/1536\n",
      "Epoch 00259: val_loss did not improve\n",
      "0s - loss: 0.0275 - acc: 0.9908 - val_loss: 0.0664 - val_acc: 0.9797\n",
      "Epoch 261/1536\n",
      "Epoch 00260: val_loss did not improve\n",
      "0s - loss: 0.0240 - acc: 0.9919 - val_loss: 0.0707 - val_acc: 0.9791\n",
      "Epoch 262/1536\n",
      "Epoch 00261: val_loss did not improve\n",
      "0s - loss: 0.0264 - acc: 0.9911 - val_loss: 0.0858 - val_acc: 0.9731\n",
      "Epoch 263/1536\n",
      "Epoch 00262: val_loss did not improve\n",
      "0s - loss: 0.3875 - acc: 0.9260 - val_loss: 0.9073 - val_acc: 0.8415\n",
      "Epoch 264/1536\n",
      "Epoch 00263: val_loss did not improve\n",
      "0s - loss: 0.3364 - acc: 0.9322 - val_loss: 0.0623 - val_acc: 0.9798\n",
      "Epoch 265/1536\n",
      "Epoch 00264: val_loss did not improve\n",
      "0s - loss: 0.0239 - acc: 0.9916 - val_loss: 0.0669 - val_acc: 0.9788\n",
      "Epoch 266/1536\n",
      "Epoch 00265: val_loss did not improve\n",
      "0s - loss: 0.0237 - acc: 0.9918 - val_loss: 0.0680 - val_acc: 0.9785\n",
      "Epoch 267/1536\n",
      "Epoch 00266: val_loss did not improve\n",
      "0s - loss: 0.0239 - acc: 0.9916 - val_loss: 0.0658 - val_acc: 0.9795\n",
      "Epoch 268/1536\n",
      "Epoch 00267: val_loss did not improve\n",
      "0s - loss: 0.1211 - acc: 0.9702 - val_loss: 0.7891 - val_acc: 0.8424\n",
      "Epoch 269/1536\n",
      "Epoch 00268: val_loss did not improve\n",
      "0s - loss: 1.3369 - acc: 0.8003 - val_loss: 0.0812 - val_acc: 0.9745\n",
      "Epoch 270/1536\n",
      "Epoch 00269: val_loss did not improve\n",
      "0s - loss: 0.0348 - acc: 0.9885 - val_loss: 0.0659 - val_acc: 0.9793\n",
      "Epoch 271/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00270: val_loss did not improve\n",
      "0s - loss: 0.0263 - acc: 0.9908 - val_loss: 0.0636 - val_acc: 0.9807\n",
      "Epoch 272/1536\n",
      "Epoch 00271: val_loss did not improve\n",
      "0s - loss: 0.0270 - acc: 0.9909 - val_loss: 0.0633 - val_acc: 0.9803\n",
      "Epoch 273/1536\n",
      "Epoch 00272: val_loss did not improve\n",
      "0s - loss: 0.0231 - acc: 0.9922 - val_loss: 0.0679 - val_acc: 0.9794\n",
      "Epoch 274/1536\n",
      "Epoch 00273: val_loss did not improve\n",
      "0s - loss: 0.0231 - acc: 0.9921 - val_loss: 0.0719 - val_acc: 0.9778\n",
      "Epoch 275/1536\n",
      "Epoch 00274: val_loss did not improve\n",
      "0s - loss: 0.6097 - acc: 0.9009 - val_loss: 1.5800 - val_acc: 0.8256\n",
      "Epoch 276/1536\n",
      "Epoch 00275: val_loss did not improve\n",
      "0s - loss: 0.8353 - acc: 0.8866 - val_loss: 0.0743 - val_acc: 0.9788\n",
      "Epoch 277/1536\n",
      "Epoch 00276: val_loss did not improve\n",
      "0s - loss: 0.0316 - acc: 0.9886 - val_loss: 0.0676 - val_acc: 0.9807\n",
      "Epoch 278/1536\n",
      "Epoch 00277: val_loss did not improve\n",
      "0s - loss: 0.0247 - acc: 0.9916 - val_loss: 0.0650 - val_acc: 0.9810\n",
      "Epoch 279/1536\n",
      "Epoch 00278: val_loss did not improve\n",
      "0s - loss: 0.0248 - acc: 0.9916 - val_loss: 0.0650 - val_acc: 0.9813\n",
      "Epoch 280/1536\n",
      "Epoch 00279: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9934 - val_loss: 0.0638 - val_acc: 0.9824\n",
      "Epoch 281/1536\n",
      "Epoch 00280: val_loss did not improve\n",
      "0s - loss: 0.0227 - acc: 0.9920 - val_loss: 0.1365 - val_acc: 0.9630\n",
      "Epoch 282/1536\n",
      "Epoch 00281: val_loss did not improve\n",
      "0s - loss: 0.7494 - acc: 0.8820 - val_loss: 2.0717 - val_acc: 0.7611\n",
      "Epoch 283/1536\n",
      "Epoch 00282: val_loss did not improve\n",
      "0s - loss: 0.8020 - acc: 0.8822 - val_loss: 0.0741 - val_acc: 0.9791\n",
      "Epoch 284/1536\n",
      "Epoch 00283: val_loss did not improve\n",
      "0s - loss: 0.0293 - acc: 0.9899 - val_loss: 0.0656 - val_acc: 0.9798\n",
      "Epoch 285/1536\n",
      "Epoch 00284: val_loss did not improve\n",
      "0s - loss: 0.0244 - acc: 0.9916 - val_loss: 0.0628 - val_acc: 0.9806\n",
      "Epoch 286/1536\n",
      "Epoch 00285: val_loss did not improve\n",
      "0s - loss: 0.0229 - acc: 0.9919 - val_loss: 0.0635 - val_acc: 0.9800\n",
      "Epoch 287/1536\n",
      "Epoch 00286: val_loss did not improve\n",
      "0s - loss: 0.0218 - acc: 0.9927 - val_loss: 0.0589 - val_acc: 0.9811\n",
      "Epoch 288/1536\n",
      "Epoch 00287: val_loss did not improve\n",
      "0s - loss: 0.0226 - acc: 0.9922 - val_loss: 0.1256 - val_acc: 0.9681\n",
      "Epoch 289/1536\n",
      "Epoch 00288: val_loss did not improve\n",
      "0s - loss: 0.5309 - acc: 0.9126 - val_loss: 0.4289 - val_acc: 0.9048\n",
      "Epoch 290/1536\n",
      "Epoch 00289: val_loss did not improve\n",
      "0s - loss: 0.2198 - acc: 0.9453 - val_loss: 0.0620 - val_acc: 0.9815\n",
      "Epoch 291/1536\n",
      "Epoch 00290: val_loss did not improve\n",
      "0s - loss: 0.0208 - acc: 0.9925 - val_loss: 0.0576 - val_acc: 0.9816\n",
      "Epoch 292/1536\n",
      "Epoch 00291: val_loss did not improve\n",
      "0s - loss: 0.0203 - acc: 0.9932 - val_loss: 0.0654 - val_acc: 0.9801\n",
      "Epoch 293/1536\n",
      "Epoch 00292: val_loss did not improve\n",
      "0s - loss: 0.0196 - acc: 0.9934 - val_loss: 0.0627 - val_acc: 0.9808\n",
      "Epoch 294/1536\n",
      "Epoch 00293: val_loss did not improve\n",
      "0s - loss: 0.0378 - acc: 0.9889 - val_loss: 0.6167 - val_acc: 0.9208\n",
      "Epoch 295/1536\n",
      "Epoch 00294: val_loss did not improve\n",
      "0s - loss: 1.0449 - acc: 0.8372 - val_loss: 0.1522 - val_acc: 0.9569\n",
      "Epoch 296/1536\n",
      "Epoch 00295: val_loss did not improve\n",
      "0s - loss: 0.0485 - acc: 0.9839 - val_loss: 0.0611 - val_acc: 0.9812\n",
      "Epoch 297/1536\n",
      "Epoch 00296: val_loss did not improve\n",
      "0s - loss: 0.0243 - acc: 0.9917 - val_loss: 0.0601 - val_acc: 0.9816\n",
      "Epoch 298/1536\n",
      "Epoch 00297: val_loss did not improve\n",
      "0s - loss: 0.0221 - acc: 0.9924 - val_loss: 0.0606 - val_acc: 0.9817\n",
      "Epoch 299/1536\n",
      "Epoch 00298: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9930 - val_loss: 0.0609 - val_acc: 0.9815\n",
      "Epoch 300/1536\n",
      "Epoch 00299: val_loss did not improve\n",
      "0s - loss: 0.0263 - acc: 0.9906 - val_loss: 0.1828 - val_acc: 0.9541\n",
      "Epoch 301/1536\n",
      "Epoch 00300: val_loss did not improve\n",
      "0s - loss: 0.6223 - acc: 0.8881 - val_loss: 1.1908 - val_acc: 0.8160\n",
      "Epoch 302/1536\n",
      "Epoch 00301: val_loss did not improve\n",
      "0s - loss: 0.5637 - acc: 0.9039 - val_loss: 0.0664 - val_acc: 0.9791\n",
      "Epoch 303/1536\n",
      "Epoch 00302: val_loss did not improve\n",
      "0s - loss: 0.0287 - acc: 0.9905 - val_loss: 0.0616 - val_acc: 0.9816\n",
      "Epoch 304/1536\n",
      "Epoch 00303: val_loss did not improve\n",
      "0s - loss: 0.0265 - acc: 0.9910 - val_loss: 0.0605 - val_acc: 0.9805\n",
      "Epoch 305/1536\n",
      "Epoch 00304: val_loss did not improve\n",
      "0s - loss: 0.0234 - acc: 0.9921 - val_loss: 0.0684 - val_acc: 0.9797\n",
      "Epoch 306/1536\n",
      "Epoch 00305: val_loss did not improve\n",
      "0s - loss: 0.0253 - acc: 0.9915 - val_loss: 0.0589 - val_acc: 0.9821\n",
      "Epoch 307/1536\n",
      "Epoch 00306: val_loss did not improve\n",
      "0s - loss: 0.0215 - acc: 0.9928 - val_loss: 0.0655 - val_acc: 0.9797\n",
      "Epoch 308/1536\n",
      "Epoch 00307: val_loss did not improve\n",
      "0s - loss: 0.7919 - acc: 0.8870 - val_loss: 3.6045 - val_acc: 0.6581\n",
      "Epoch 309/1536\n",
      "Epoch 00308: val_loss did not improve\n",
      "0s - loss: 1.5348 - acc: 0.8240 - val_loss: 0.0765 - val_acc: 0.9775\n",
      "Epoch 310/1536\n",
      "Epoch 00309: val_loss did not improve\n",
      "0s - loss: 0.0349 - acc: 0.9879 - val_loss: 0.0674 - val_acc: 0.9809\n",
      "Epoch 311/1536\n",
      "Epoch 00310: val_loss did not improve\n",
      "0s - loss: 0.0312 - acc: 0.9889 - val_loss: 0.0654 - val_acc: 0.9806\n",
      "Epoch 312/1536\n",
      "Epoch 00311: val_loss did not improve\n",
      "0s - loss: 0.0271 - acc: 0.9904 - val_loss: 0.0645 - val_acc: 0.9815\n",
      "Epoch 313/1536\n",
      "Epoch 00312: val_loss did not improve\n",
      "0s - loss: 0.0268 - acc: 0.9906 - val_loss: 0.0636 - val_acc: 0.9815\n",
      "Epoch 314/1536\n",
      "Epoch 00313: val_loss did not improve\n",
      "0s - loss: 0.0232 - acc: 0.9915 - val_loss: 0.0604 - val_acc: 0.9814\n",
      "Epoch 315/1536\n",
      "Epoch 00314: val_loss did not improve\n",
      "0s - loss: 0.0219 - acc: 0.9927 - val_loss: 0.0933 - val_acc: 0.9739\n",
      "Epoch 316/1536\n",
      "Epoch 00315: val_loss did not improve\n",
      "0s - loss: 0.5767 - acc: 0.9051 - val_loss: 0.3118 - val_acc: 0.9329\n",
      "Epoch 317/1536\n",
      "Epoch 00316: val_loss did not improve\n",
      "0s - loss: 0.2987 - acc: 0.9427 - val_loss: 0.0650 - val_acc: 0.9800\n",
      "Epoch 318/1536\n",
      "Epoch 00317: val_loss did not improve\n",
      "0s - loss: 0.0231 - acc: 0.9925 - val_loss: 0.0597 - val_acc: 0.9820\n",
      "Epoch 319/1536\n",
      "Epoch 00318: val_loss did not improve\n",
      "0s - loss: 0.0222 - acc: 0.9929 - val_loss: 0.0583 - val_acc: 0.9820\n",
      "Epoch 320/1536\n",
      "Epoch 00319: val_loss did not improve\n",
      "0s - loss: 0.0207 - acc: 0.9935 - val_loss: 0.0750 - val_acc: 0.9777\n",
      "Epoch 321/1536\n",
      "Epoch 00320: val_loss did not improve\n",
      "0s - loss: 0.3305 - acc: 0.9400 - val_loss: 1.2221 - val_acc: 0.8483\n",
      "Epoch 322/1536\n",
      "Epoch 00321: val_loss did not improve\n",
      "0s - loss: 0.9483 - acc: 0.8649 - val_loss: 0.0663 - val_acc: 0.9807\n",
      "Epoch 323/1536\n",
      "Epoch 00322: val_loss did not improve\n",
      "0s - loss: 0.0250 - acc: 0.9916 - val_loss: 0.0652 - val_acc: 0.9807\n",
      "Epoch 324/1536\n",
      "Epoch 00323: val_loss did not improve\n",
      "0s - loss: 0.0250 - acc: 0.9914 - val_loss: 0.0631 - val_acc: 0.9808\n",
      "Epoch 325/1536\n",
      "Epoch 00324: val_loss did not improve\n",
      "0s - loss: 0.0234 - acc: 0.9916 - val_loss: 0.0711 - val_acc: 0.9792\n",
      "Epoch 326/1536\n",
      "Epoch 00325: val_loss did not improve\n",
      "0s - loss: 0.0242 - acc: 0.9916 - val_loss: 0.0873 - val_acc: 0.9745\n",
      "Epoch 327/1536\n",
      "Epoch 00326: val_loss did not improve\n",
      "0s - loss: 0.2829 - acc: 0.9327 - val_loss: 0.3974 - val_acc: 0.9115\n",
      "Epoch 328/1536\n",
      "Epoch 00327: val_loss did not improve\n",
      "0s - loss: 0.2786 - acc: 0.9294 - val_loss: 0.0686 - val_acc: 0.9807\n",
      "Epoch 329/1536\n",
      "Epoch 00328: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9932 - val_loss: 0.0622 - val_acc: 0.9808\n",
      "Epoch 330/1536\n",
      "Epoch 00329: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9940 - val_loss: 0.0639 - val_acc: 0.9815\n",
      "Epoch 331/1536\n",
      "Epoch 00330: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9939 - val_loss: 0.0573 - val_acc: 0.9825\n",
      "Epoch 332/1536\n",
      "Epoch 00331: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9947 - val_loss: 0.0820 - val_acc: 0.9753\n",
      "Epoch 333/1536\n",
      "Epoch 00332: val_loss did not improve\n",
      "0s - loss: 0.6827 - acc: 0.8885 - val_loss: 1.1118 - val_acc: 0.8205\n",
      "Epoch 334/1536\n",
      "Epoch 00333: val_loss did not improve\n",
      "0s - loss: 0.4500 - acc: 0.9158 - val_loss: 0.0690 - val_acc: 0.9788\n",
      "Epoch 335/1536\n",
      "Epoch 00334: val_loss did not improve\n",
      "0s - loss: 0.0253 - acc: 0.9911 - val_loss: 0.0657 - val_acc: 0.9802\n",
      "Epoch 336/1536\n",
      "Epoch 00335: val_loss did not improve\n",
      "0s - loss: 0.0251 - acc: 0.9912 - val_loss: 0.0679 - val_acc: 0.9791\n",
      "Epoch 337/1536\n",
      "Epoch 00336: val_loss did not improve\n",
      "0s - loss: 0.0228 - acc: 0.9916 - val_loss: 0.0623 - val_acc: 0.9796\n",
      "Epoch 338/1536\n",
      "Epoch 00337: val_loss did not improve\n",
      "0s - loss: 0.0254 - acc: 0.9914 - val_loss: 0.4188 - val_acc: 0.9192\n",
      "Epoch 339/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00338: val_loss did not improve\n",
      "0s - loss: 0.5963 - acc: 0.8869 - val_loss: 0.3769 - val_acc: 0.9202\n",
      "Epoch 340/1536\n",
      "Epoch 00339: val_loss did not improve\n",
      "0s - loss: 0.1676 - acc: 0.9583 - val_loss: 0.0600 - val_acc: 0.9808\n",
      "Epoch 341/1536\n",
      "Epoch 00340: val_loss did not improve\n",
      "0s - loss: 0.0211 - acc: 0.9930 - val_loss: 0.0568 - val_acc: 0.9809\n",
      "Epoch 342/1536\n",
      "Epoch 00341: val_loss did not improve\n",
      "0s - loss: 0.0200 - acc: 0.9928 - val_loss: 0.0606 - val_acc: 0.9816\n",
      "Epoch 343/1536\n",
      "Epoch 00342: val_loss did not improve\n",
      "0s - loss: 0.0181 - acc: 0.9943 - val_loss: 0.0557 - val_acc: 0.9826\n",
      "Epoch 344/1536\n",
      "Epoch 00343: val_loss did not improve\n",
      "0s - loss: 0.0207 - acc: 0.9933 - val_loss: 0.0725 - val_acc: 0.9790\n",
      "Epoch 345/1536\n",
      "Epoch 00344: val_loss did not improve\n",
      "0s - loss: 0.3611 - acc: 0.9295 - val_loss: 0.6484 - val_acc: 0.8802\n",
      "Epoch 346/1536\n",
      "Epoch 00345: val_loss did not improve\n",
      "0s - loss: 0.3042 - acc: 0.9299 - val_loss: 0.0583 - val_acc: 0.9825\n",
      "Epoch 347/1536\n",
      "Epoch 00346: val_loss did not improve\n",
      "0s - loss: 0.0200 - acc: 0.9934 - val_loss: 0.0553 - val_acc: 0.9824\n",
      "Epoch 348/1536\n",
      "Epoch 00347: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9944 - val_loss: 0.0614 - val_acc: 0.9800\n",
      "Epoch 349/1536\n",
      "Epoch 00348: val_loss did not improve\n",
      "0s - loss: 0.0200 - acc: 0.9931 - val_loss: 0.0546 - val_acc: 0.9822\n",
      "Epoch 350/1536\n",
      "Epoch 00349: val_loss did not improve\n",
      "0s - loss: 0.0458 - acc: 0.9882 - val_loss: 0.7443 - val_acc: 0.8951\n",
      "Epoch 351/1536\n",
      "Epoch 00350: val_loss did not improve\n",
      "0s - loss: 1.2275 - acc: 0.8172 - val_loss: 0.3535 - val_acc: 0.9195\n",
      "Epoch 352/1536\n",
      "Epoch 00351: val_loss did not improve\n",
      "0s - loss: 0.1091 - acc: 0.9708 - val_loss: 0.0673 - val_acc: 0.9794\n",
      "Epoch 353/1536\n",
      "Epoch 00352: val_loss did not improve\n",
      "0s - loss: 0.0249 - acc: 0.9915 - val_loss: 0.0626 - val_acc: 0.9808\n",
      "Epoch 354/1536\n",
      "Epoch 00353: val_loss did not improve\n",
      "0s - loss: 0.0235 - acc: 0.9925 - val_loss: 0.0637 - val_acc: 0.9816\n",
      "Epoch 355/1536\n",
      "Epoch 00354: val_loss did not improve\n",
      "0s - loss: 0.0220 - acc: 0.9928 - val_loss: 0.0629 - val_acc: 0.9818\n",
      "Epoch 356/1536\n",
      "Epoch 00355: val_loss did not improve\n",
      "0s - loss: 0.0207 - acc: 0.9930 - val_loss: 0.0655 - val_acc: 0.9804\n",
      "Epoch 357/1536\n",
      "Epoch 00356: val_loss did not improve\n",
      "0s - loss: 0.0267 - acc: 0.9905 - val_loss: 0.2825 - val_acc: 0.9346\n",
      "Epoch 358/1536\n",
      "Epoch 00357: val_loss did not improve\n",
      "0s - loss: 0.6412 - acc: 0.8943 - val_loss: 0.8383 - val_acc: 0.8403\n",
      "Epoch 359/1536\n",
      "Epoch 00358: val_loss did not improve\n",
      "0s - loss: 0.5289 - acc: 0.9015 - val_loss: 0.0627 - val_acc: 0.9821\n",
      "Epoch 360/1536\n",
      "Epoch 00359: val_loss did not improve\n",
      "0s - loss: 0.0246 - acc: 0.9918 - val_loss: 0.0578 - val_acc: 0.9832\n",
      "Epoch 361/1536\n",
      "Epoch 00360: val_loss did not improve\n",
      "0s - loss: 0.0219 - acc: 0.9929 - val_loss: 0.0575 - val_acc: 0.9836\n",
      "Epoch 362/1536\n",
      "Epoch 00361: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9934 - val_loss: 0.0587 - val_acc: 0.9823\n",
      "Epoch 363/1536\n",
      "Epoch 00362: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9930 - val_loss: 0.0648 - val_acc: 0.9811\n",
      "Epoch 364/1536\n",
      "Epoch 00363: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9935 - val_loss: 0.0704 - val_acc: 0.9788\n",
      "Epoch 365/1536\n",
      "Epoch 00364: val_loss did not improve\n",
      "0s - loss: 0.3314 - acc: 0.9441 - val_loss: 3.1868 - val_acc: 0.7461\n",
      "Epoch 366/1536\n",
      "Epoch 00365: val_loss did not improve\n",
      "0s - loss: 1.7066 - acc: 0.7943 - val_loss: 0.0751 - val_acc: 0.9773\n",
      "Epoch 367/1536\n",
      "Epoch 00366: val_loss did not improve\n",
      "0s - loss: 0.0310 - acc: 0.9891 - val_loss: 0.0618 - val_acc: 0.9811\n",
      "Epoch 368/1536\n",
      "Epoch 00367: val_loss did not improve\n",
      "0s - loss: 0.0264 - acc: 0.9909 - val_loss: 0.0603 - val_acc: 0.9808\n",
      "Epoch 369/1536\n",
      "Epoch 00368: val_loss did not improve\n",
      "0s - loss: 0.0234 - acc: 0.9924 - val_loss: 0.0580 - val_acc: 0.9814\n",
      "Epoch 370/1536\n",
      "Epoch 00369: val_loss did not improve\n",
      "0s - loss: 0.0211 - acc: 0.9929 - val_loss: 0.0579 - val_acc: 0.9821\n",
      "Epoch 371/1536\n",
      "Epoch 00370: val_loss did not improve\n",
      "0s - loss: 0.0173 - acc: 0.9941 - val_loss: 0.0575 - val_acc: 0.9831\n",
      "Epoch 372/1536\n",
      "Epoch 00371: val_loss did not improve\n",
      "0s - loss: 0.0203 - acc: 0.9929 - val_loss: 0.0828 - val_acc: 0.9745\n",
      "Epoch 373/1536\n",
      "Epoch 00372: val_loss did not improve\n",
      "0s - loss: 0.2218 - acc: 0.9574 - val_loss: 0.7552 - val_acc: 0.8650\n",
      "Epoch 374/1536\n",
      "Epoch 00373: val_loss did not improve\n",
      "0s - loss: 0.6092 - acc: 0.8975 - val_loss: 0.0641 - val_acc: 0.9806\n",
      "Epoch 375/1536\n",
      "Epoch 00374: val_loss did not improve\n",
      "0s - loss: 0.0233 - acc: 0.9924 - val_loss: 0.0630 - val_acc: 0.9819\n",
      "Epoch 376/1536\n",
      "Epoch 00375: val_loss did not improve\n",
      "0s - loss: 0.0202 - acc: 0.9932 - val_loss: 0.0626 - val_acc: 0.9813\n",
      "Epoch 377/1536\n",
      "Epoch 00376: val_loss did not improve\n",
      "0s - loss: 0.0209 - acc: 0.9929 - val_loss: 0.0705 - val_acc: 0.9794\n",
      "Epoch 378/1536\n",
      "Epoch 00377: val_loss did not improve\n",
      "0s - loss: 0.0605 - acc: 0.9818 - val_loss: 0.4991 - val_acc: 0.9073\n",
      "Epoch 379/1536\n",
      "Epoch 00378: val_loss did not improve\n",
      "0s - loss: 0.4110 - acc: 0.9103 - val_loss: 0.3610 - val_acc: 0.9128\n",
      "Epoch 380/1536\n",
      "Epoch 00379: val_loss did not improve\n",
      "0s - loss: 0.1698 - acc: 0.9558 - val_loss: 0.0586 - val_acc: 0.9816\n",
      "Epoch 381/1536\n",
      "Epoch 00380: val_loss did not improve\n",
      "0s - loss: 0.0189 - acc: 0.9937 - val_loss: 0.0563 - val_acc: 0.9825\n",
      "Epoch 382/1536\n",
      "Epoch 00381: val_loss did not improve\n",
      "0s - loss: 0.0180 - acc: 0.9940 - val_loss: 0.0579 - val_acc: 0.9817\n",
      "Epoch 383/1536\n",
      "Epoch 00382: val_loss did not improve\n",
      "0s - loss: 0.0192 - acc: 0.9937 - val_loss: 0.0683 - val_acc: 0.9775\n",
      "Epoch 384/1536\n",
      "Epoch 00383: val_loss did not improve\n",
      "0s - loss: 0.5424 - acc: 0.9235 - val_loss: 0.9003 - val_acc: 0.8647\n",
      "Epoch 385/1536\n",
      "Epoch 00384: val_loss did not improve\n",
      "0s - loss: 0.2968 - acc: 0.9456 - val_loss: 0.0620 - val_acc: 0.9805\n",
      "Epoch 386/1536\n",
      "Epoch 00385: val_loss did not improve\n",
      "0s - loss: 0.0221 - acc: 0.9929 - val_loss: 0.0608 - val_acc: 0.9822\n",
      "Epoch 387/1536\n",
      "Epoch 00386: val_loss did not improve\n",
      "0s - loss: 0.0221 - acc: 0.9927 - val_loss: 0.0620 - val_acc: 0.9813\n",
      "Epoch 388/1536\n",
      "Epoch 00387: val_loss did not improve\n",
      "0s - loss: 0.0228 - acc: 0.9924 - val_loss: 0.0630 - val_acc: 0.9800\n",
      "Epoch 389/1536\n",
      "Epoch 00388: val_loss did not improve\n",
      "0s - loss: 0.1337 - acc: 0.9706 - val_loss: 0.2338 - val_acc: 0.9453\n",
      "Epoch 390/1536\n",
      "Epoch 00389: val_loss did not improve\n",
      "0s - loss: 0.9042 - acc: 0.8668 - val_loss: 0.6278 - val_acc: 0.8861\n",
      "Epoch 391/1536\n",
      "Epoch 00390: val_loss did not improve\n",
      "0s - loss: 0.1809 - acc: 0.9628 - val_loss: 0.0641 - val_acc: 0.9823\n",
      "Epoch 392/1536\n",
      "Epoch 00391: val_loss did not improve\n",
      "0s - loss: 0.0235 - acc: 0.9922 - val_loss: 0.0633 - val_acc: 0.9820\n",
      "Epoch 393/1536\n",
      "Epoch 00392: val_loss did not improve\n",
      "0s - loss: 0.0211 - acc: 0.9932 - val_loss: 0.0606 - val_acc: 0.9825\n",
      "Epoch 394/1536\n",
      "Epoch 00393: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9934 - val_loss: 0.0632 - val_acc: 0.9808\n",
      "Epoch 395/1536\n",
      "Epoch 00394: val_loss did not improve\n",
      "0s - loss: 0.0323 - acc: 0.9886 - val_loss: 0.1502 - val_acc: 0.9633\n",
      "Epoch 396/1536\n",
      "Epoch 00395: val_loss did not improve\n",
      "0s - loss: 0.3725 - acc: 0.9299 - val_loss: 0.8525 - val_acc: 0.8791\n",
      "Epoch 397/1536\n",
      "Epoch 00396: val_loss did not improve\n",
      "0s - loss: 0.2659 - acc: 0.9473 - val_loss: 0.0551 - val_acc: 0.9833\n",
      "Epoch 398/1536\n",
      "Epoch 00397: val_loss did not improve\n",
      "0s - loss: 0.0186 - acc: 0.9937 - val_loss: 0.0550 - val_acc: 0.9817\n",
      "Epoch 399/1536\n",
      "Epoch 00398: val_loss did not improve\n",
      "0s - loss: 0.0165 - acc: 0.9946 - val_loss: 0.0557 - val_acc: 0.9825\n",
      "Epoch 400/1536\n",
      "Epoch 00399: val_loss improved from 0.05442 to 0.05408, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0162 - acc: 0.9949 - val_loss: 0.0541 - val_acc: 0.9831\n",
      "Epoch 401/1536\n",
      "Epoch 00400: val_loss did not improve\n",
      "0s - loss: 0.0175 - acc: 0.9942 - val_loss: 0.1028 - val_acc: 0.9700\n",
      "Epoch 402/1536\n",
      "Epoch 00401: val_loss did not improve\n",
      "0s - loss: 0.5457 - acc: 0.8971 - val_loss: 1.1881 - val_acc: 0.7974\n",
      "Epoch 403/1536\n",
      "Epoch 00402: val_loss did not improve\n",
      "0s - loss: 0.8498 - acc: 0.8608 - val_loss: 0.0760 - val_acc: 0.9779\n",
      "Epoch 404/1536\n",
      "Epoch 00403: val_loss did not improve\n",
      "0s - loss: 0.0284 - acc: 0.9900 - val_loss: 0.0682 - val_acc: 0.9803\n",
      "Epoch 405/1536\n",
      "Epoch 00404: val_loss did not improve\n",
      "0s - loss: 0.0257 - acc: 0.9910 - val_loss: 0.0659 - val_acc: 0.9798\n",
      "Epoch 406/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00405: val_loss did not improve\n",
      "0s - loss: 0.0223 - acc: 0.9918 - val_loss: 0.0687 - val_acc: 0.9797\n",
      "Epoch 407/1536\n",
      "Epoch 00406: val_loss did not improve\n",
      "0s - loss: 0.0217 - acc: 0.9925 - val_loss: 0.0703 - val_acc: 0.9800\n",
      "Epoch 408/1536\n",
      "Epoch 00407: val_loss did not improve\n",
      "0s - loss: 0.0201 - acc: 0.9929 - val_loss: 0.0683 - val_acc: 0.9804\n",
      "Epoch 409/1536\n",
      "Epoch 00408: val_loss did not improve\n",
      "0s - loss: 0.5833 - acc: 0.9196 - val_loss: 0.8031 - val_acc: 0.8539\n",
      "Epoch 410/1536\n",
      "Epoch 00409: val_loss did not improve\n",
      "0s - loss: 0.4012 - acc: 0.9150 - val_loss: 0.0654 - val_acc: 0.9808\n",
      "Epoch 411/1536\n",
      "Epoch 00410: val_loss did not improve\n",
      "0s - loss: 0.0218 - acc: 0.9925 - val_loss: 0.0635 - val_acc: 0.9809\n",
      "Epoch 412/1536\n",
      "Epoch 00411: val_loss did not improve\n",
      "0s - loss: 0.0201 - acc: 0.9928 - val_loss: 0.0624 - val_acc: 0.9808\n",
      "Epoch 413/1536\n",
      "Epoch 00412: val_loss did not improve\n",
      "0s - loss: 0.0207 - acc: 0.9926 - val_loss: 0.0603 - val_acc: 0.9818\n",
      "Epoch 414/1536\n",
      "Epoch 00413: val_loss did not improve\n",
      "0s - loss: 0.0180 - acc: 0.9935 - val_loss: 0.0663 - val_acc: 0.9808\n",
      "Epoch 415/1536\n",
      "Epoch 00414: val_loss did not improve\n",
      "0s - loss: 0.1720 - acc: 0.9678 - val_loss: 0.3366 - val_acc: 0.9205\n",
      "Epoch 416/1536\n",
      "Epoch 00415: val_loss did not improve\n",
      "0s - loss: 1.0853 - acc: 0.8318 - val_loss: 0.1632 - val_acc: 0.9593\n",
      "Epoch 417/1536\n",
      "Epoch 00416: val_loss did not improve\n",
      "0s - loss: 0.0428 - acc: 0.9865 - val_loss: 0.0629 - val_acc: 0.9818\n",
      "Epoch 418/1536\n",
      "Epoch 00417: val_loss did not improve\n",
      "0s - loss: 0.0209 - acc: 0.9929 - val_loss: 0.0632 - val_acc: 0.9822\n",
      "Epoch 419/1536\n",
      "Epoch 00418: val_loss did not improve\n",
      "0s - loss: 0.0202 - acc: 0.9931 - val_loss: 0.0627 - val_acc: 0.9820\n",
      "Epoch 420/1536\n",
      "Epoch 00419: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9933 - val_loss: 0.0608 - val_acc: 0.9821\n",
      "Epoch 421/1536\n",
      "Epoch 00420: val_loss did not improve\n",
      "0s - loss: 0.0278 - acc: 0.9901 - val_loss: 0.1461 - val_acc: 0.9571\n",
      "Epoch 422/1536\n",
      "Epoch 00421: val_loss did not improve\n",
      "0s - loss: 0.4339 - acc: 0.9148 - val_loss: 1.2107 - val_acc: 0.7957\n",
      "Epoch 423/1536\n",
      "Epoch 00422: val_loss did not improve\n",
      "0s - loss: 0.9761 - acc: 0.8448 - val_loss: 0.0716 - val_acc: 0.9797\n",
      "Epoch 424/1536\n",
      "Epoch 00423: val_loss did not improve\n",
      "0s - loss: 0.0257 - acc: 0.9914 - val_loss: 0.0599 - val_acc: 0.9820\n",
      "Epoch 425/1536\n",
      "Epoch 00424: val_loss did not improve\n",
      "0s - loss: 0.0223 - acc: 0.9924 - val_loss: 0.0585 - val_acc: 0.9832\n",
      "Epoch 426/1536\n",
      "Epoch 00425: val_loss did not improve\n",
      "0s - loss: 0.0178 - acc: 0.9938 - val_loss: 0.0561 - val_acc: 0.9833\n",
      "Epoch 427/1536\n",
      "Epoch 00426: val_loss did not improve\n",
      "0s - loss: 0.0183 - acc: 0.9944 - val_loss: 0.0549 - val_acc: 0.9831\n",
      "Epoch 428/1536\n",
      "Epoch 00427: val_loss did not improve\n",
      "0s - loss: 0.0159 - acc: 0.9948 - val_loss: 0.0578 - val_acc: 0.9828\n",
      "Epoch 429/1536\n",
      "Epoch 00428: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9941 - val_loss: 0.0568 - val_acc: 0.9822\n",
      "Epoch 430/1536\n",
      "Epoch 00429: val_loss did not improve\n",
      "0s - loss: 0.1925 - acc: 0.9650 - val_loss: 1.6352 - val_acc: 0.7999\n",
      "Epoch 431/1536\n",
      "Epoch 00430: val_loss did not improve\n",
      "0s - loss: 1.4696 - acc: 0.8004 - val_loss: 0.1035 - val_acc: 0.9716\n",
      "Epoch 432/1536\n",
      "Epoch 00431: val_loss did not improve\n",
      "0s - loss: 0.0378 - acc: 0.9876 - val_loss: 0.0671 - val_acc: 0.9804\n",
      "Epoch 433/1536\n",
      "Epoch 00432: val_loss did not improve\n",
      "0s - loss: 0.0229 - acc: 0.9917 - val_loss: 0.0626 - val_acc: 0.9825\n",
      "Epoch 434/1536\n",
      "Epoch 00433: val_loss did not improve\n",
      "0s - loss: 0.0194 - acc: 0.9932 - val_loss: 0.0643 - val_acc: 0.9817\n",
      "Epoch 435/1536\n",
      "Epoch 00434: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9933 - val_loss: 0.0591 - val_acc: 0.9828\n",
      "Epoch 436/1536\n",
      "Epoch 00435: val_loss did not improve\n",
      "0s - loss: 0.0181 - acc: 0.9941 - val_loss: 0.0746 - val_acc: 0.9791\n",
      "Epoch 437/1536\n",
      "Epoch 00436: val_loss did not improve\n",
      "0s - loss: 0.1059 - acc: 0.9783 - val_loss: 0.1824 - val_acc: 0.9539\n",
      "Epoch 438/1536\n",
      "Epoch 00437: val_loss did not improve\n",
      "0s - loss: 0.0921 - acc: 0.9719 - val_loss: 0.0604 - val_acc: 0.9813\n",
      "Epoch 439/1536\n",
      "Epoch 00438: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9944 - val_loss: 0.0572 - val_acc: 0.9813\n",
      "Epoch 440/1536\n",
      "Epoch 00439: val_loss did not improve\n",
      "0s - loss: 0.0607 - acc: 0.9836 - val_loss: 0.5222 - val_acc: 0.8973\n",
      "Epoch 441/1536\n",
      "Epoch 00440: val_loss did not improve\n",
      "0s - loss: 0.8362 - acc: 0.8641 - val_loss: 0.3760 - val_acc: 0.9165\n",
      "Epoch 442/1536\n",
      "Epoch 00441: val_loss did not improve\n",
      "0s - loss: 0.1363 - acc: 0.9645 - val_loss: 0.0589 - val_acc: 0.9828\n",
      "Epoch 443/1536\n",
      "Epoch 00442: val_loss did not improve\n",
      "0s - loss: 0.0193 - acc: 0.9933 - val_loss: 0.0566 - val_acc: 0.9828\n",
      "Epoch 444/1536\n",
      "Epoch 00443: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9938 - val_loss: 0.0575 - val_acc: 0.9833\n",
      "Epoch 445/1536\n",
      "Epoch 00444: val_loss did not improve\n",
      "0s - loss: 0.0162 - acc: 0.9948 - val_loss: 0.0556 - val_acc: 0.9833\n",
      "Epoch 446/1536\n",
      "Epoch 00445: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9952 - val_loss: 0.0665 - val_acc: 0.9799\n",
      "Epoch 447/1536\n",
      "Epoch 00446: val_loss did not improve\n",
      "0s - loss: 0.5385 - acc: 0.9185 - val_loss: 0.7322 - val_acc: 0.8816\n",
      "Epoch 448/1536\n",
      "Epoch 00447: val_loss did not improve\n",
      "0s - loss: 0.4323 - acc: 0.9265 - val_loss: 0.0647 - val_acc: 0.9801\n",
      "Epoch 449/1536\n",
      "Epoch 00448: val_loss did not improve\n",
      "0s - loss: 0.0244 - acc: 0.9915 - val_loss: 0.0595 - val_acc: 0.9817\n",
      "Epoch 450/1536\n",
      "Epoch 00449: val_loss did not improve\n",
      "0s - loss: 0.0203 - acc: 0.9932 - val_loss: 0.0590 - val_acc: 0.9837\n",
      "Epoch 451/1536\n",
      "Epoch 00450: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9935 - val_loss: 0.0689 - val_acc: 0.9802\n",
      "Epoch 452/1536\n",
      "Epoch 00451: val_loss did not improve\n",
      "0s - loss: 0.0249 - acc: 0.9909 - val_loss: 0.0656 - val_acc: 0.9814\n",
      "Epoch 453/1536\n",
      "Epoch 00452: val_loss did not improve\n",
      "0s - loss: 0.1760 - acc: 0.9627 - val_loss: 0.6811 - val_acc: 0.8855\n",
      "Epoch 454/1536\n",
      "Epoch 00453: val_loss did not improve\n",
      "0s - loss: 0.4725 - acc: 0.8991 - val_loss: 0.2773 - val_acc: 0.9388\n",
      "Epoch 455/1536\n",
      "Epoch 00454: val_loss did not improve\n",
      "0s - loss: 0.1063 - acc: 0.9741 - val_loss: 0.0547 - val_acc: 0.9821\n",
      "Epoch 456/1536\n",
      "Epoch 00455: val_loss improved from 0.05408 to 0.05381, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0170 - acc: 0.9949 - val_loss: 0.0538 - val_acc: 0.9825\n",
      "Epoch 457/1536\n",
      "Epoch 00456: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9950 - val_loss: 0.0539 - val_acc: 0.9828\n",
      "Epoch 458/1536\n",
      "Epoch 00457: val_loss improved from 0.05381 to 0.05171, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0154 - acc: 0.9950 - val_loss: 0.0517 - val_acc: 0.9825\n",
      "Epoch 459/1536\n",
      "Epoch 00458: val_loss did not improve\n",
      "0s - loss: 0.0169 - acc: 0.9946 - val_loss: 0.0778 - val_acc: 0.9745\n",
      "Epoch 460/1536\n",
      "Epoch 00459: val_loss did not improve\n",
      "0s - loss: 0.5800 - acc: 0.9039 - val_loss: 1.3649 - val_acc: 0.8185\n",
      "Epoch 461/1536\n",
      "Epoch 00460: val_loss did not improve\n",
      "0s - loss: 0.8289 - acc: 0.8803 - val_loss: 0.0691 - val_acc: 0.9788\n",
      "Epoch 462/1536\n",
      "Epoch 00461: val_loss did not improve\n",
      "0s - loss: 0.0264 - acc: 0.9913 - val_loss: 0.0649 - val_acc: 0.9808\n",
      "Epoch 463/1536\n",
      "Epoch 00462: val_loss did not improve\n",
      "0s - loss: 0.0229 - acc: 0.9925 - val_loss: 0.0607 - val_acc: 0.9821\n",
      "Epoch 464/1536\n",
      "Epoch 00463: val_loss did not improve\n",
      "0s - loss: 0.0213 - acc: 0.9925 - val_loss: 0.0610 - val_acc: 0.9810\n",
      "Epoch 465/1536\n",
      "Epoch 00464: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9937 - val_loss: 0.0756 - val_acc: 0.9776\n",
      "Epoch 466/1536\n",
      "Epoch 00465: val_loss did not improve\n",
      "0s - loss: 0.0259 - acc: 0.9910 - val_loss: 0.0648 - val_acc: 0.9808\n",
      "Epoch 467/1536\n",
      "Epoch 00466: val_loss did not improve\n",
      "0s - loss: 0.0567 - acc: 0.9842 - val_loss: 0.4247 - val_acc: 0.9134\n",
      "Epoch 468/1536\n",
      "Epoch 00467: val_loss did not improve\n",
      "0s - loss: 0.6174 - acc: 0.8869 - val_loss: 0.5349 - val_acc: 0.8903\n",
      "Epoch 469/1536\n",
      "Epoch 00468: val_loss did not improve\n",
      "0s - loss: 0.2496 - acc: 0.9474 - val_loss: 0.0595 - val_acc: 0.9825\n",
      "Epoch 470/1536\n",
      "Epoch 00469: val_loss did not improve\n",
      "0s - loss: 0.0204 - acc: 0.9933 - val_loss: 0.0578 - val_acc: 0.9831\n",
      "Epoch 471/1536\n",
      "Epoch 00470: val_loss did not improve\n",
      "0s - loss: 0.0175 - acc: 0.9941 - val_loss: 0.0607 - val_acc: 0.9817\n",
      "Epoch 472/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00471: val_loss did not improve\n",
      "0s - loss: 0.0180 - acc: 0.9945 - val_loss: 0.0586 - val_acc: 0.9825\n",
      "Epoch 473/1536\n",
      "Epoch 00472: val_loss did not improve\n",
      "0s - loss: 0.0165 - acc: 0.9948 - val_loss: 0.0582 - val_acc: 0.9817\n",
      "Epoch 474/1536\n",
      "Epoch 00473: val_loss did not improve\n",
      "0s - loss: 0.0268 - acc: 0.9906 - val_loss: 0.4073 - val_acc: 0.9280\n",
      "Epoch 475/1536\n",
      "Epoch 00474: val_loss did not improve\n",
      "0s - loss: 0.7514 - acc: 0.8696 - val_loss: 0.6524 - val_acc: 0.8744\n",
      "Epoch 476/1536\n",
      "Epoch 00475: val_loss did not improve\n",
      "0s - loss: 0.2067 - acc: 0.9556 - val_loss: 0.0643 - val_acc: 0.9800\n",
      "Epoch 477/1536\n",
      "Epoch 00476: val_loss did not improve\n",
      "0s - loss: 0.0243 - acc: 0.9915 - val_loss: 0.0592 - val_acc: 0.9820\n",
      "Epoch 478/1536\n",
      "Epoch 00477: val_loss did not improve\n",
      "0s - loss: 0.0221 - acc: 0.9930 - val_loss: 0.0692 - val_acc: 0.9783\n",
      "Epoch 479/1536\n",
      "Epoch 00478: val_loss did not improve\n",
      "0s - loss: 0.0223 - acc: 0.9927 - val_loss: 0.0569 - val_acc: 0.9829\n",
      "Epoch 480/1536\n",
      "Epoch 00479: val_loss did not improve\n",
      "0s - loss: 0.0209 - acc: 0.9930 - val_loss: 0.0568 - val_acc: 0.9837\n",
      "Epoch 481/1536\n",
      "Epoch 00480: val_loss did not improve\n",
      "0s - loss: 0.1210 - acc: 0.9707 - val_loss: 1.3757 - val_acc: 0.7693\n",
      "Epoch 482/1536\n",
      "Epoch 00481: val_loss did not improve\n",
      "0s - loss: 1.6601 - acc: 0.7680 - val_loss: 0.1062 - val_acc: 0.9704\n",
      "Epoch 483/1536\n",
      "Epoch 00482: val_loss did not improve\n",
      "0s - loss: 0.0415 - acc: 0.9862 - val_loss: 0.0661 - val_acc: 0.9806\n",
      "Epoch 484/1536\n",
      "Epoch 00483: val_loss did not improve\n",
      "0s - loss: 0.0253 - acc: 0.9908 - val_loss: 0.0625 - val_acc: 0.9817\n",
      "Epoch 485/1536\n",
      "Epoch 00484: val_loss did not improve\n",
      "0s - loss: 0.0226 - acc: 0.9919 - val_loss: 0.0632 - val_acc: 0.9817\n",
      "Epoch 486/1536\n",
      "Epoch 00485: val_loss did not improve\n",
      "0s - loss: 0.0207 - acc: 0.9929 - val_loss: 0.0616 - val_acc: 0.9812\n",
      "Epoch 487/1536\n",
      "Epoch 00486: val_loss did not improve\n",
      "0s - loss: 0.0195 - acc: 0.9933 - val_loss: 0.0599 - val_acc: 0.9820\n",
      "Epoch 488/1536\n",
      "Epoch 00487: val_loss did not improve\n",
      "0s - loss: 0.0178 - acc: 0.9938 - val_loss: 0.0721 - val_acc: 0.9783\n",
      "Epoch 489/1536\n",
      "Epoch 00488: val_loss did not improve\n",
      "0s - loss: 0.4077 - acc: 0.9283 - val_loss: 1.0444 - val_acc: 0.8549\n",
      "Epoch 490/1536\n",
      "Epoch 00489: val_loss did not improve\n",
      "0s - loss: 0.3654 - acc: 0.9357 - val_loss: 0.0572 - val_acc: 0.9826\n",
      "Epoch 491/1536\n",
      "Epoch 00490: val_loss did not improve\n",
      "0s - loss: 0.0188 - acc: 0.9936 - val_loss: 0.0565 - val_acc: 0.9826\n",
      "Epoch 492/1536\n",
      "Epoch 00491: val_loss did not improve\n",
      "0s - loss: 0.0157 - acc: 0.9950 - val_loss: 0.0575 - val_acc: 0.9818\n",
      "Epoch 493/1536\n",
      "Epoch 00492: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9943 - val_loss: 0.0553 - val_acc: 0.9825\n",
      "Epoch 494/1536\n",
      "Epoch 00493: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9954 - val_loss: 0.0912 - val_acc: 0.9736\n",
      "Epoch 495/1536\n",
      "Epoch 00494: val_loss did not improve\n",
      "0s - loss: 0.6494 - acc: 0.8894 - val_loss: 0.7861 - val_acc: 0.9082\n",
      "Epoch 496/1536\n",
      "Epoch 00495: val_loss did not improve\n",
      "0s - loss: 0.3500 - acc: 0.9377 - val_loss: 0.0640 - val_acc: 0.9811\n",
      "Epoch 497/1536\n",
      "Epoch 00496: val_loss did not improve\n",
      "0s - loss: 0.0196 - acc: 0.9933 - val_loss: 0.0622 - val_acc: 0.9813\n",
      "Epoch 498/1536\n",
      "Epoch 00497: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9935 - val_loss: 0.0587 - val_acc: 0.9825\n",
      "Epoch 499/1536\n",
      "Epoch 00498: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9936 - val_loss: 0.0653 - val_acc: 0.9807\n",
      "Epoch 500/1536\n",
      "Epoch 00499: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9944 - val_loss: 0.0622 - val_acc: 0.9816\n",
      "Epoch 501/1536\n",
      "Epoch 00500: val_loss did not improve\n",
      "0s - loss: 0.0232 - acc: 0.9918 - val_loss: 0.2009 - val_acc: 0.9504\n",
      "Epoch 502/1536\n",
      "Epoch 00501: val_loss did not improve\n",
      "0s - loss: 1.1202 - acc: 0.8413 - val_loss: 0.1459 - val_acc: 0.9627\n",
      "Epoch 503/1536\n",
      "Epoch 00502: val_loss did not improve\n",
      "0s - loss: 0.0422 - acc: 0.9858 - val_loss: 0.0650 - val_acc: 0.9804\n",
      "Epoch 504/1536\n",
      "Epoch 00503: val_loss did not improve\n",
      "0s - loss: 0.0235 - acc: 0.9916 - val_loss: 0.0648 - val_acc: 0.9808\n",
      "Epoch 505/1536\n",
      "Epoch 00504: val_loss did not improve\n",
      "0s - loss: 0.0211 - acc: 0.9928 - val_loss: 0.0665 - val_acc: 0.9813\n",
      "Epoch 506/1536\n",
      "Epoch 00505: val_loss did not improve\n",
      "0s - loss: 0.0175 - acc: 0.9937 - val_loss: 0.0641 - val_acc: 0.9821\n",
      "Epoch 507/1536\n",
      "Epoch 00506: val_loss did not improve\n",
      "0s - loss: 0.0189 - acc: 0.9934 - val_loss: 0.0964 - val_acc: 0.9736\n",
      "Epoch 508/1536\n",
      "Epoch 00507: val_loss did not improve\n",
      "0s - loss: 1.0910 - acc: 0.8611 - val_loss: 0.4516 - val_acc: 0.9085\n",
      "Epoch 509/1536\n",
      "Epoch 00508: val_loss did not improve\n",
      "0s - loss: 0.1481 - acc: 0.9672 - val_loss: 0.0691 - val_acc: 0.9810\n",
      "Epoch 510/1536\n",
      "Epoch 00509: val_loss did not improve\n",
      "0s - loss: 0.0204 - acc: 0.9929 - val_loss: 0.0657 - val_acc: 0.9822\n",
      "Epoch 511/1536\n",
      "Epoch 00510: val_loss did not improve\n",
      "0s - loss: 0.0178 - acc: 0.9936 - val_loss: 0.0621 - val_acc: 0.9829\n",
      "Epoch 512/1536\n",
      "Epoch 00511: val_loss did not improve\n",
      "0s - loss: 0.0190 - acc: 0.9930 - val_loss: 0.0740 - val_acc: 0.9784\n",
      "Epoch 513/1536\n",
      "Epoch 00512: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9945 - val_loss: 0.0682 - val_acc: 0.9827\n",
      "Epoch 514/1536\n",
      "Epoch 00513: val_loss did not improve\n",
      "0s - loss: 0.0415 - acc: 0.9870 - val_loss: 0.3499 - val_acc: 0.9314\n",
      "Epoch 515/1536\n",
      "Epoch 00514: val_loss did not improve\n",
      "0s - loss: 1.0093 - acc: 0.8517 - val_loss: 0.9726 - val_acc: 0.8587\n",
      "Epoch 516/1536\n",
      "Epoch 00515: val_loss did not improve\n",
      "0s - loss: 0.3535 - acc: 0.9411 - val_loss: 0.0657 - val_acc: 0.9808\n",
      "Epoch 517/1536\n",
      "Epoch 00516: val_loss did not improve\n",
      "0s - loss: 0.0210 - acc: 0.9928 - val_loss: 0.0608 - val_acc: 0.9825\n",
      "Epoch 518/1536\n",
      "Epoch 00517: val_loss did not improve\n",
      "0s - loss: 0.0191 - acc: 0.9935 - val_loss: 0.0609 - val_acc: 0.9824\n",
      "Epoch 519/1536\n",
      "Epoch 00518: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9942 - val_loss: 0.0618 - val_acc: 0.9823\n",
      "Epoch 520/1536\n",
      "Epoch 00519: val_loss did not improve\n",
      "0s - loss: 0.0165 - acc: 0.9944 - val_loss: 0.0579 - val_acc: 0.9839\n",
      "Epoch 521/1536\n",
      "Epoch 00520: val_loss did not improve\n",
      "0s - loss: 0.0184 - acc: 0.9933 - val_loss: 0.0664 - val_acc: 0.9814\n",
      "Epoch 522/1536\n",
      "Epoch 00521: val_loss did not improve\n",
      "0s - loss: 0.1520 - acc: 0.9737 - val_loss: 0.7256 - val_acc: 0.8781\n",
      "Epoch 523/1536\n",
      "Epoch 00522: val_loss did not improve\n",
      "0s - loss: 0.7759 - acc: 0.8715 - val_loss: 0.5022 - val_acc: 0.9099\n",
      "Epoch 524/1536\n",
      "Epoch 00523: val_loss did not improve\n",
      "0s - loss: 0.1501 - acc: 0.9678 - val_loss: 0.0623 - val_acc: 0.9828\n",
      "Epoch 525/1536\n",
      "Epoch 00524: val_loss did not improve\n",
      "0s - loss: 0.0171 - acc: 0.9940 - val_loss: 0.0583 - val_acc: 0.9831\n",
      "Epoch 526/1536\n",
      "Epoch 00525: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9948 - val_loss: 0.0592 - val_acc: 0.9828\n",
      "Epoch 527/1536\n",
      "Epoch 00526: val_loss did not improve\n",
      "0s - loss: 0.0154 - acc: 0.9946 - val_loss: 0.0588 - val_acc: 0.9816\n",
      "Epoch 528/1536\n",
      "Epoch 00527: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9955 - val_loss: 0.0573 - val_acc: 0.9823\n",
      "Epoch 529/1536\n",
      "Epoch 00528: val_loss did not improve\n",
      "0s - loss: 0.0662 - acc: 0.9825 - val_loss: 0.4705 - val_acc: 0.9157\n",
      "Epoch 530/1536\n",
      "Epoch 00529: val_loss did not improve\n",
      "0s - loss: 0.5084 - acc: 0.9050 - val_loss: 0.2124 - val_acc: 0.9490\n",
      "Epoch 531/1536\n",
      "Epoch 00530: val_loss did not improve\n",
      "0s - loss: 0.0951 - acc: 0.9747 - val_loss: 0.0609 - val_acc: 0.9823\n",
      "Epoch 532/1536\n",
      "Epoch 00531: val_loss did not improve\n",
      "0s - loss: 0.0178 - acc: 0.9940 - val_loss: 0.0611 - val_acc: 0.9825\n",
      "Epoch 533/1536\n",
      "Epoch 00532: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9947 - val_loss: 0.0681 - val_acc: 0.9810\n",
      "Epoch 534/1536\n",
      "Epoch 00533: val_loss did not improve\n",
      "0s - loss: 0.0181 - acc: 0.9941 - val_loss: 0.1090 - val_acc: 0.9699\n",
      "Epoch 535/1536\n",
      "Epoch 00534: val_loss did not improve\n",
      "0s - loss: 0.7528 - acc: 0.8877 - val_loss: 0.1886 - val_acc: 0.9511\n",
      "Epoch 536/1536\n",
      "Epoch 00535: val_loss did not improve\n",
      "0s - loss: 0.0516 - acc: 0.9835 - val_loss: 0.0626 - val_acc: 0.9816\n",
      "Epoch 537/1536\n",
      "Epoch 00536: val_loss did not improve\n",
      "0s - loss: 0.0213 - acc: 0.9928 - val_loss: 0.0634 - val_acc: 0.9808\n",
      "Epoch 538/1536\n",
      "Epoch 00537: val_loss did not improve\n",
      "0s - loss: 0.0174 - acc: 0.9944 - val_loss: 0.0624 - val_acc: 0.9810\n",
      "Epoch 539/1536\n",
      "Epoch 00538: val_loss did not improve\n",
      "0s - loss: 0.0308 - acc: 0.9894 - val_loss: 0.3984 - val_acc: 0.9314\n",
      "Epoch 540/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00539: val_loss did not improve\n",
      "0s - loss: 0.7497 - acc: 0.8792 - val_loss: 0.5870 - val_acc: 0.8824\n",
      "Epoch 541/1536\n",
      "Epoch 00540: val_loss did not improve\n",
      "0s - loss: 0.2050 - acc: 0.9509 - val_loss: 0.0568 - val_acc: 0.9819\n",
      "Epoch 542/1536\n",
      "Epoch 00541: val_loss did not improve\n",
      "0s - loss: 0.0186 - acc: 0.9937 - val_loss: 0.0564 - val_acc: 0.9827\n",
      "Epoch 543/1536\n",
      "Epoch 00542: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9945 - val_loss: 0.0537 - val_acc: 0.9828\n",
      "Epoch 544/1536\n",
      "Epoch 00543: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9948 - val_loss: 0.0577 - val_acc: 0.9817\n",
      "Epoch 545/1536\n",
      "Epoch 00544: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9950 - val_loss: 0.0641 - val_acc: 0.9807\n",
      "Epoch 546/1536\n",
      "Epoch 00545: val_loss did not improve\n",
      "0s - loss: 0.0216 - acc: 0.9926 - val_loss: 0.4043 - val_acc: 0.9341\n",
      "Epoch 547/1536\n",
      "Epoch 00546: val_loss did not improve\n",
      "0s - loss: 1.0676 - acc: 0.8429 - val_loss: 1.0293 - val_acc: 0.8567\n",
      "Epoch 548/1536\n",
      "Epoch 00547: val_loss did not improve\n",
      "0s - loss: 0.2412 - acc: 0.9584 - val_loss: 0.0642 - val_acc: 0.9798\n",
      "Epoch 549/1536\n",
      "Epoch 00548: val_loss did not improve\n",
      "0s - loss: 0.0231 - acc: 0.9921 - val_loss: 0.0600 - val_acc: 0.9818\n",
      "Epoch 550/1536\n",
      "Epoch 00549: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9935 - val_loss: 0.0606 - val_acc: 0.9808\n",
      "Epoch 551/1536\n",
      "Epoch 00550: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9936 - val_loss: 0.0597 - val_acc: 0.9808\n",
      "Epoch 552/1536\n",
      "Epoch 00551: val_loss did not improve\n",
      "0s - loss: 0.0179 - acc: 0.9943 - val_loss: 0.0599 - val_acc: 0.9810\n",
      "Epoch 553/1536\n",
      "Epoch 00552: val_loss did not improve\n",
      "0s - loss: 0.0205 - acc: 0.9933 - val_loss: 0.0781 - val_acc: 0.9767\n",
      "Epoch 554/1536\n",
      "Epoch 00553: val_loss did not improve\n",
      "0s - loss: 0.5146 - acc: 0.9147 - val_loss: 2.6812 - val_acc: 0.6999\n",
      "Epoch 555/1536\n",
      "Epoch 00554: val_loss did not improve\n",
      "0s - loss: 0.9753 - acc: 0.8684 - val_loss: 0.0644 - val_acc: 0.9808\n",
      "Epoch 556/1536\n",
      "Epoch 00555: val_loss did not improve\n",
      "0s - loss: 0.0284 - acc: 0.9907 - val_loss: 0.0613 - val_acc: 0.9820\n",
      "Epoch 557/1536\n",
      "Epoch 00556: val_loss did not improve\n",
      "0s - loss: 0.0233 - acc: 0.9919 - val_loss: 0.0578 - val_acc: 0.9833\n",
      "Epoch 558/1536\n",
      "Epoch 00557: val_loss did not improve\n",
      "0s - loss: 0.0202 - acc: 0.9930 - val_loss: 0.0559 - val_acc: 0.9836\n",
      "Epoch 559/1536\n",
      "Epoch 00558: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9942 - val_loss: 0.0577 - val_acc: 0.9832\n",
      "Epoch 560/1536\n",
      "Epoch 00559: val_loss did not improve\n",
      "0s - loss: 0.0165 - acc: 0.9945 - val_loss: 0.0700 - val_acc: 0.9792\n",
      "Epoch 561/1536\n",
      "Epoch 00560: val_loss did not improve\n",
      "0s - loss: 0.2078 - acc: 0.9554 - val_loss: 0.9127 - val_acc: 0.8633\n",
      "Epoch 562/1536\n",
      "Epoch 00561: val_loss did not improve\n",
      "0s - loss: 0.7369 - acc: 0.8831 - val_loss: 0.0632 - val_acc: 0.9809\n",
      "Epoch 563/1536\n",
      "Epoch 00562: val_loss did not improve\n",
      "0s - loss: 0.0200 - acc: 0.9935 - val_loss: 0.0575 - val_acc: 0.9822\n",
      "Epoch 564/1536\n",
      "Epoch 00563: val_loss did not improve\n",
      "0s - loss: 0.0189 - acc: 0.9935 - val_loss: 0.0633 - val_acc: 0.9812\n",
      "Epoch 565/1536\n",
      "Epoch 00564: val_loss did not improve\n",
      "0s - loss: 0.0172 - acc: 0.9943 - val_loss: 0.0603 - val_acc: 0.9829\n",
      "Epoch 566/1536\n",
      "Epoch 00565: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9943 - val_loss: 0.0611 - val_acc: 0.9824\n",
      "Epoch 567/1536\n",
      "Epoch 00566: val_loss did not improve\n",
      "0s - loss: 0.0209 - acc: 0.9926 - val_loss: 0.1555 - val_acc: 0.9576\n",
      "Epoch 568/1536\n",
      "Epoch 00567: val_loss did not improve\n",
      "0s - loss: 0.7296 - acc: 0.8877 - val_loss: 1.4961 - val_acc: 0.7864\n",
      "Epoch 569/1536\n",
      "Epoch 00568: val_loss did not improve\n",
      "0s - loss: 0.4920 - acc: 0.9206 - val_loss: 0.0693 - val_acc: 0.9812\n",
      "Epoch 570/1536\n",
      "Epoch 00569: val_loss did not improve\n",
      "0s - loss: 0.0235 - acc: 0.9926 - val_loss: 0.0663 - val_acc: 0.9825\n",
      "Epoch 571/1536\n",
      "Epoch 00570: val_loss did not improve\n",
      "0s - loss: 0.0183 - acc: 0.9939 - val_loss: 0.0622 - val_acc: 0.9825\n",
      "Epoch 572/1536\n",
      "Epoch 00571: val_loss did not improve\n",
      "0s - loss: 0.0169 - acc: 0.9941 - val_loss: 0.0617 - val_acc: 0.9825\n",
      "Epoch 573/1536\n",
      "Epoch 00572: val_loss did not improve\n",
      "0s - loss: 0.0152 - acc: 0.9953 - val_loss: 0.0636 - val_acc: 0.9825\n",
      "Epoch 574/1536\n",
      "Epoch 00573: val_loss did not improve\n",
      "0s - loss: 0.0162 - acc: 0.9943 - val_loss: 0.0566 - val_acc: 0.9844\n",
      "Epoch 575/1536\n",
      "Epoch 00574: val_loss did not improve\n",
      "0s - loss: 0.0197 - acc: 0.9931 - val_loss: 0.3800 - val_acc: 0.9232\n",
      "Epoch 576/1536\n",
      "Epoch 00575: val_loss did not improve\n",
      "0s - loss: 0.8368 - acc: 0.8634 - val_loss: 1.2903 - val_acc: 0.8314\n",
      "Epoch 577/1536\n",
      "Epoch 00576: val_loss did not improve\n",
      "0s - loss: 0.4853 - acc: 0.9163 - val_loss: 0.0619 - val_acc: 0.9824\n",
      "Epoch 578/1536\n",
      "Epoch 00577: val_loss did not improve\n",
      "0s - loss: 0.0220 - acc: 0.9923 - val_loss: 0.0607 - val_acc: 0.9821\n",
      "Epoch 579/1536\n",
      "Epoch 00578: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9928 - val_loss: 0.0596 - val_acc: 0.9829\n",
      "Epoch 580/1536\n",
      "Epoch 00579: val_loss did not improve\n",
      "0s - loss: 0.0191 - acc: 0.9929 - val_loss: 0.0590 - val_acc: 0.9824\n",
      "Epoch 581/1536\n",
      "Epoch 00580: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9938 - val_loss: 0.0620 - val_acc: 0.9826\n",
      "Epoch 582/1536\n",
      "Epoch 00581: val_loss did not improve\n",
      "0s - loss: 0.0183 - acc: 0.9936 - val_loss: 0.1378 - val_acc: 0.9627\n",
      "Epoch 583/1536\n",
      "Epoch 00582: val_loss did not improve\n",
      "0s - loss: 0.5524 - acc: 0.8972 - val_loss: 1.1827 - val_acc: 0.8612\n",
      "Epoch 584/1536\n",
      "Epoch 00583: val_loss did not improve\n",
      "0s - loss: 0.5439 - acc: 0.9221 - val_loss: 0.0661 - val_acc: 0.9811\n",
      "Epoch 585/1536\n",
      "Epoch 00584: val_loss did not improve\n",
      "0s - loss: 0.0242 - acc: 0.9919 - val_loss: 0.0636 - val_acc: 0.9824\n",
      "Epoch 586/1536\n",
      "Epoch 00585: val_loss did not improve\n",
      "0s - loss: 0.0216 - acc: 0.9931 - val_loss: 0.0607 - val_acc: 0.9828\n",
      "Epoch 587/1536\n",
      "Epoch 00586: val_loss did not improve\n",
      "0s - loss: 0.0169 - acc: 0.9943 - val_loss: 0.0629 - val_acc: 0.9820\n",
      "Epoch 588/1536\n",
      "Epoch 00587: val_loss did not improve\n",
      "0s - loss: 0.0168 - acc: 0.9944 - val_loss: 0.0634 - val_acc: 0.9823\n",
      "Epoch 589/1536\n",
      "Epoch 00588: val_loss did not improve\n",
      "0s - loss: 0.0174 - acc: 0.9942 - val_loss: 0.0830 - val_acc: 0.9770\n",
      "Epoch 590/1536\n",
      "Epoch 00589: val_loss did not improve\n",
      "0s - loss: 1.4711 - acc: 0.8478 - val_loss: 2.4680 - val_acc: 0.7434\n",
      "Epoch 591/1536\n",
      "Epoch 00590: val_loss did not improve\n",
      "0s - loss: 1.1141 - acc: 0.8646 - val_loss: 0.0772 - val_acc: 0.9786\n",
      "Epoch 592/1536\n",
      "Epoch 00591: val_loss did not improve\n",
      "0s - loss: 0.0405 - acc: 0.9866 - val_loss: 0.0675 - val_acc: 0.9815\n",
      "Epoch 593/1536\n",
      "Epoch 00592: val_loss did not improve\n",
      "0s - loss: 0.0294 - acc: 0.9897 - val_loss: 0.0658 - val_acc: 0.9813\n",
      "Epoch 594/1536\n",
      "Epoch 00593: val_loss did not improve\n",
      "0s - loss: 0.0251 - acc: 0.9912 - val_loss: 0.0610 - val_acc: 0.9828\n",
      "Epoch 595/1536\n",
      "Epoch 00594: val_loss did not improve\n",
      "0s - loss: 0.0222 - acc: 0.9926 - val_loss: 0.0625 - val_acc: 0.9825\n",
      "Epoch 596/1536\n",
      "Epoch 00595: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9926 - val_loss: 0.0587 - val_acc: 0.9836\n",
      "Epoch 597/1536\n",
      "Epoch 00596: val_loss did not improve\n",
      "0s - loss: 0.0194 - acc: 0.9929 - val_loss: 0.0967 - val_acc: 0.9755\n",
      "Epoch 598/1536\n",
      "Epoch 00597: val_loss did not improve\n",
      "0s - loss: 0.5348 - acc: 0.9098 - val_loss: 0.3333 - val_acc: 0.9273\n",
      "Epoch 599/1536\n",
      "Epoch 00598: val_loss did not improve\n",
      "0s - loss: 0.1452 - acc: 0.9650 - val_loss: 0.0564 - val_acc: 0.9836\n",
      "Epoch 600/1536\n",
      "Epoch 00599: val_loss did not improve\n",
      "0s - loss: 0.0161 - acc: 0.9945 - val_loss: 0.0548 - val_acc: 0.9848\n",
      "Epoch 601/1536\n",
      "Epoch 00600: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9948 - val_loss: 0.0646 - val_acc: 0.9816\n",
      "Epoch 602/1536\n",
      "Epoch 00601: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9946 - val_loss: 0.0552 - val_acc: 0.9842\n",
      "Epoch 603/1536\n",
      "Epoch 00602: val_loss did not improve\n",
      "0s - loss: 0.0239 - acc: 0.9918 - val_loss: 0.3987 - val_acc: 0.9359\n",
      "Epoch 604/1536\n",
      "Epoch 00603: val_loss did not improve\n",
      "0s - loss: 0.7801 - acc: 0.8842 - val_loss: 0.3364 - val_acc: 0.9377\n",
      "Epoch 605/1536\n",
      "Epoch 00604: val_loss did not improve\n",
      "0s - loss: 0.0823 - acc: 0.9810 - val_loss: 0.0640 - val_acc: 0.9825\n",
      "Epoch 606/1536\n",
      "Epoch 00605: val_loss did not improve\n",
      "0s - loss: 0.0190 - acc: 0.9932 - val_loss: 0.0692 - val_acc: 0.9800\n",
      "Epoch 607/1536\n",
      "Epoch 00606: val_loss did not improve\n",
      "0s - loss: 0.0180 - acc: 0.9937 - val_loss: 0.0626 - val_acc: 0.9820\n",
      "Epoch 608/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00607: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9947 - val_loss: 0.0562 - val_acc: 0.9847\n",
      "Epoch 609/1536\n",
      "Epoch 00608: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9950 - val_loss: 0.0576 - val_acc: 0.9841\n",
      "Epoch 610/1536\n",
      "Epoch 00609: val_loss did not improve\n",
      "0s - loss: 0.2814 - acc: 0.9457 - val_loss: 0.9728 - val_acc: 0.8728\n",
      "Epoch 611/1536\n",
      "Epoch 00610: val_loss did not improve\n",
      "0s - loss: 0.4841 - acc: 0.9130 - val_loss: 0.0597 - val_acc: 0.9816\n",
      "Epoch 612/1536\n",
      "Epoch 00611: val_loss did not improve\n",
      "0s - loss: 0.0172 - acc: 0.9941 - val_loss: 0.0606 - val_acc: 0.9811\n",
      "Epoch 613/1536\n",
      "Epoch 00612: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9942 - val_loss: 0.0546 - val_acc: 0.9841\n",
      "Epoch 614/1536\n",
      "Epoch 00613: val_loss did not improve\n",
      "0s - loss: 0.0151 - acc: 0.9951 - val_loss: 0.0523 - val_acc: 0.9846\n",
      "Epoch 615/1536\n",
      "Epoch 00614: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9950 - val_loss: 0.0582 - val_acc: 0.9839\n",
      "Epoch 616/1536\n",
      "Epoch 00615: val_loss did not improve\n",
      "0s - loss: 0.4566 - acc: 0.9344 - val_loss: 0.8962 - val_acc: 0.8829\n",
      "Epoch 617/1536\n",
      "Epoch 00616: val_loss did not improve\n",
      "0s - loss: 0.7429 - acc: 0.8923 - val_loss: 0.0563 - val_acc: 0.9832\n",
      "Epoch 618/1536\n",
      "Epoch 00617: val_loss did not improve\n",
      "0s - loss: 0.0233 - acc: 0.9920 - val_loss: 0.0559 - val_acc: 0.9835\n",
      "Epoch 619/1536\n",
      "Epoch 00618: val_loss did not improve\n",
      "0s - loss: 0.0192 - acc: 0.9934 - val_loss: 0.0536 - val_acc: 0.9832\n",
      "Epoch 620/1536\n",
      "Epoch 00619: val_loss improved from 0.05171 to 0.05169, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0178 - acc: 0.9944 - val_loss: 0.0517 - val_acc: 0.9841\n",
      "Epoch 621/1536\n",
      "Epoch 00620: val_loss did not improve\n",
      "0s - loss: 0.0151 - acc: 0.9948 - val_loss: 0.0545 - val_acc: 0.9837\n",
      "Epoch 622/1536\n",
      "Epoch 00621: val_loss did not improve\n",
      "0s - loss: 0.0183 - acc: 0.9937 - val_loss: 0.0537 - val_acc: 0.9841\n",
      "Epoch 623/1536\n",
      "Epoch 00622: val_loss did not improve\n",
      "0s - loss: 0.1965 - acc: 0.9592 - val_loss: 0.5027 - val_acc: 0.9046\n",
      "Epoch 624/1536\n",
      "Epoch 00623: val_loss did not improve\n",
      "0s - loss: 0.2263 - acc: 0.9459 - val_loss: 0.0673 - val_acc: 0.9816\n",
      "Epoch 625/1536\n",
      "Epoch 00624: val_loss did not improve\n",
      "0s - loss: 0.0182 - acc: 0.9937 - val_loss: 0.0602 - val_acc: 0.9829\n",
      "Epoch 626/1536\n",
      "Epoch 00625: val_loss did not improve\n",
      "0s - loss: 0.0153 - acc: 0.9949 - val_loss: 0.0576 - val_acc: 0.9829\n",
      "Epoch 627/1536\n",
      "Epoch 00626: val_loss did not improve\n",
      "0s - loss: 0.0143 - acc: 0.9952 - val_loss: 0.0595 - val_acc: 0.9822\n",
      "Epoch 628/1536\n",
      "Epoch 00627: val_loss did not improve\n",
      "0s - loss: 0.3436 - acc: 0.9399 - val_loss: 1.7568 - val_acc: 0.7911\n",
      "Epoch 629/1536\n",
      "Epoch 00628: val_loss did not improve\n",
      "0s - loss: 0.5452 - acc: 0.9120 - val_loss: 0.0609 - val_acc: 0.9815\n",
      "Epoch 630/1536\n",
      "Epoch 00629: val_loss did not improve\n",
      "0s - loss: 0.0201 - acc: 0.9932 - val_loss: 0.0592 - val_acc: 0.9832\n",
      "Epoch 631/1536\n",
      "Epoch 00630: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9940 - val_loss: 0.0532 - val_acc: 0.9845\n",
      "Epoch 632/1536\n",
      "Epoch 00631: val_loss did not improve\n",
      "0s - loss: 0.0156 - acc: 0.9943 - val_loss: 0.0568 - val_acc: 0.9847\n",
      "Epoch 633/1536\n",
      "Epoch 00632: val_loss did not improve\n",
      "0s - loss: 0.0157 - acc: 0.9946 - val_loss: 0.0561 - val_acc: 0.9847\n",
      "Epoch 634/1536\n",
      "Epoch 00633: val_loss did not improve\n",
      "0s - loss: 0.0297 - acc: 0.9907 - val_loss: 0.9413 - val_acc: 0.8880\n",
      "Epoch 635/1536\n",
      "Epoch 00634: val_loss did not improve\n",
      "0s - loss: 0.8615 - acc: 0.8677 - val_loss: 0.8296 - val_acc: 0.8692\n",
      "Epoch 636/1536\n",
      "Epoch 00635: val_loss did not improve\n",
      "0s - loss: 0.2870 - acc: 0.9443 - val_loss: 0.0631 - val_acc: 0.9830\n",
      "Epoch 637/1536\n",
      "Epoch 00636: val_loss did not improve\n",
      "0s - loss: 0.0191 - acc: 0.9937 - val_loss: 0.0631 - val_acc: 0.9827\n",
      "Epoch 638/1536\n",
      "Epoch 00637: val_loss did not improve\n",
      "0s - loss: 0.0154 - acc: 0.9946 - val_loss: 0.0589 - val_acc: 0.9841\n",
      "Epoch 639/1536\n",
      "Epoch 00638: val_loss did not improve\n",
      "0s - loss: 0.0141 - acc: 0.9957 - val_loss: 0.0619 - val_acc: 0.9825\n",
      "Epoch 640/1536\n",
      "Epoch 00639: val_loss did not improve\n",
      "0s - loss: 0.0148 - acc: 0.9945 - val_loss: 0.0631 - val_acc: 0.9828\n",
      "Epoch 641/1536\n",
      "Epoch 00640: val_loss did not improve\n",
      "0s - loss: 0.0136 - acc: 0.9955 - val_loss: 0.0589 - val_acc: 0.9834\n",
      "Epoch 642/1536\n",
      "Epoch 00641: val_loss did not improve\n",
      "0s - loss: 0.0159 - acc: 0.9945 - val_loss: 0.6879 - val_acc: 0.8867\n",
      "Epoch 643/1536\n",
      "Epoch 00642: val_loss did not improve\n",
      "0s - loss: 0.7488 - acc: 0.8814 - val_loss: 0.1174 - val_acc: 0.9709\n",
      "Epoch 644/1536\n",
      "Epoch 00643: val_loss did not improve\n",
      "0s - loss: 0.0269 - acc: 0.9915 - val_loss: 0.0626 - val_acc: 0.9822\n",
      "Epoch 645/1536\n",
      "Epoch 00644: val_loss did not improve\n",
      "0s - loss: 0.0172 - acc: 0.9941 - val_loss: 0.0589 - val_acc: 0.9834\n",
      "Epoch 646/1536\n",
      "Epoch 00645: val_loss did not improve\n",
      "0s - loss: 0.0151 - acc: 0.9950 - val_loss: 0.0592 - val_acc: 0.9841\n",
      "Epoch 647/1536\n",
      "Epoch 00646: val_loss did not improve\n",
      "0s - loss: 0.0229 - acc: 0.9921 - val_loss: 0.5800 - val_acc: 0.9081\n",
      "Epoch 648/1536\n",
      "Epoch 00647: val_loss did not improve\n",
      "0s - loss: 0.7613 - acc: 0.8655 - val_loss: 0.4713 - val_acc: 0.9067\n",
      "Epoch 649/1536\n",
      "Epoch 00648: val_loss did not improve\n",
      "0s - loss: 0.1376 - acc: 0.9697 - val_loss: 0.0609 - val_acc: 0.9820\n",
      "Epoch 650/1536\n",
      "Epoch 00649: val_loss did not improve\n",
      "0s - loss: 0.0179 - acc: 0.9939 - val_loss: 0.0622 - val_acc: 0.9807\n",
      "Epoch 651/1536\n",
      "Epoch 00650: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9943 - val_loss: 0.0595 - val_acc: 0.9833\n",
      "Epoch 652/1536\n",
      "Epoch 00651: val_loss did not improve\n",
      "0s - loss: 0.0153 - acc: 0.9945 - val_loss: 0.0576 - val_acc: 0.9832\n",
      "Epoch 653/1536\n",
      "Epoch 00652: val_loss did not improve\n",
      "0s - loss: 0.0154 - acc: 0.9948 - val_loss: 0.0566 - val_acc: 0.9835\n",
      "Epoch 654/1536\n",
      "Epoch 00653: val_loss did not improve\n",
      "0s - loss: 0.0199 - acc: 0.9933 - val_loss: 0.5677 - val_acc: 0.8988\n",
      "Epoch 655/1536\n",
      "Epoch 00654: val_loss did not improve\n",
      "0s - loss: 1.7653 - acc: 0.7998 - val_loss: 0.3415 - val_acc: 0.9358\n",
      "Epoch 656/1536\n",
      "Epoch 00655: val_loss did not improve\n",
      "0s - loss: 0.0925 - acc: 0.9773 - val_loss: 0.0682 - val_acc: 0.9799\n",
      "Epoch 657/1536\n",
      "Epoch 00656: val_loss did not improve\n",
      "0s - loss: 0.0246 - acc: 0.9916 - val_loss: 0.0643 - val_acc: 0.9812\n",
      "Epoch 658/1536\n",
      "Epoch 00657: val_loss did not improve\n",
      "0s - loss: 0.0201 - acc: 0.9931 - val_loss: 0.0626 - val_acc: 0.9816\n",
      "Epoch 659/1536\n",
      "Epoch 00658: val_loss did not improve\n",
      "0s - loss: 0.0184 - acc: 0.9940 - val_loss: 0.0620 - val_acc: 0.9823\n",
      "Epoch 660/1536\n",
      "Epoch 00659: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9938 - val_loss: 0.0588 - val_acc: 0.9834\n",
      "Epoch 661/1536\n",
      "Epoch 00660: val_loss did not improve\n",
      "0s - loss: 0.0151 - acc: 0.9946 - val_loss: 0.0719 - val_acc: 0.9797\n",
      "Epoch 662/1536\n",
      "Epoch 00661: val_loss did not improve\n",
      "0s - loss: 0.0663 - acc: 0.9808 - val_loss: 0.4307 - val_acc: 0.9107\n",
      "Epoch 663/1536\n",
      "Epoch 00662: val_loss did not improve\n",
      "0s - loss: 0.7833 - acc: 0.8765 - val_loss: 0.0737 - val_acc: 0.9798\n",
      "Epoch 664/1536\n",
      "Epoch 00663: val_loss did not improve\n",
      "0s - loss: 0.0200 - acc: 0.9931 - val_loss: 0.0624 - val_acc: 0.9823\n",
      "Epoch 665/1536\n",
      "Epoch 00664: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9947 - val_loss: 0.0635 - val_acc: 0.9827\n",
      "Epoch 666/1536\n",
      "Epoch 00665: val_loss did not improve\n",
      "0s - loss: 0.0144 - acc: 0.9949 - val_loss: 0.0611 - val_acc: 0.9825\n",
      "Epoch 667/1536\n",
      "Epoch 00666: val_loss did not improve\n",
      "0s - loss: 0.0139 - acc: 0.9955 - val_loss: 0.0645 - val_acc: 0.9820\n",
      "Epoch 668/1536\n",
      "Epoch 00667: val_loss did not improve\n",
      "0s - loss: 0.0145 - acc: 0.9949 - val_loss: 0.1762 - val_acc: 0.9552\n",
      "Epoch 669/1536\n",
      "Epoch 00668: val_loss did not improve\n",
      "0s - loss: 0.6274 - acc: 0.8893 - val_loss: 0.5998 - val_acc: 0.8853\n",
      "Epoch 670/1536\n",
      "Epoch 00669: val_loss did not improve\n",
      "0s - loss: 0.1617 - acc: 0.9641 - val_loss: 0.0702 - val_acc: 0.9808\n",
      "Epoch 671/1536\n",
      "Epoch 00670: val_loss did not improve\n",
      "0s - loss: 0.0191 - acc: 0.9937 - val_loss: 0.0694 - val_acc: 0.9802\n",
      "Epoch 672/1536\n",
      "Epoch 00671: val_loss did not improve\n",
      "0s - loss: 0.0168 - acc: 0.9941 - val_loss: 0.0629 - val_acc: 0.9818\n",
      "Epoch 673/1536\n",
      "Epoch 00672: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9941 - val_loss: 0.0878 - val_acc: 0.9771\n",
      "Epoch 674/1536\n",
      "Epoch 00673: val_loss did not improve\n",
      "0s - loss: 0.3358 - acc: 0.9370 - val_loss: 0.3998 - val_acc: 0.9196\n",
      "Epoch 675/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00674: val_loss did not improve\n",
      "0s - loss: 0.6127 - acc: 0.8924 - val_loss: 0.0875 - val_acc: 0.9760\n",
      "Epoch 676/1536\n",
      "Epoch 00675: val_loss did not improve\n",
      "0s - loss: 0.0220 - acc: 0.9928 - val_loss: 0.0652 - val_acc: 0.9826\n",
      "Epoch 677/1536\n",
      "Epoch 00676: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9949 - val_loss: 0.0629 - val_acc: 0.9823\n",
      "Epoch 678/1536\n",
      "Epoch 00677: val_loss did not improve\n",
      "0s - loss: 0.0147 - acc: 0.9951 - val_loss: 0.0651 - val_acc: 0.9810\n",
      "Epoch 679/1536\n",
      "Epoch 00678: val_loss did not improve\n",
      "0s - loss: 0.0159 - acc: 0.9946 - val_loss: 0.0640 - val_acc: 0.9818\n",
      "Epoch 680/1536\n",
      "Epoch 00679: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9954 - val_loss: 0.0603 - val_acc: 0.9819\n",
      "Epoch 681/1536\n",
      "Epoch 00680: val_loss did not improve\n",
      "0s - loss: 0.0215 - acc: 0.9930 - val_loss: 0.5990 - val_acc: 0.9309\n",
      "Epoch 682/1536\n",
      "Epoch 00681: val_loss did not improve\n",
      "0s - loss: 1.0180 - acc: 0.8641 - val_loss: 1.0738 - val_acc: 0.8440\n",
      "Epoch 683/1536\n",
      "Epoch 00682: val_loss did not improve\n",
      "0s - loss: 0.4653 - acc: 0.9224 - val_loss: 0.0639 - val_acc: 0.9820\n",
      "Epoch 684/1536\n",
      "Epoch 00683: val_loss did not improve\n",
      "0s - loss: 0.0215 - acc: 0.9924 - val_loss: 0.0597 - val_acc: 0.9827\n",
      "Epoch 685/1536\n",
      "Epoch 00684: val_loss did not improve\n",
      "0s - loss: 0.0193 - acc: 0.9932 - val_loss: 0.0575 - val_acc: 0.9832\n",
      "Epoch 686/1536\n",
      "Epoch 00685: val_loss did not improve\n",
      "0s - loss: 0.0172 - acc: 0.9944 - val_loss: 0.0566 - val_acc: 0.9837\n",
      "Epoch 687/1536\n",
      "Epoch 00686: val_loss did not improve\n",
      "0s - loss: 0.0143 - acc: 0.9951 - val_loss: 0.0552 - val_acc: 0.9841\n",
      "Epoch 688/1536\n",
      "Epoch 00687: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9947 - val_loss: 0.0671 - val_acc: 0.9803\n",
      "Epoch 689/1536\n",
      "Epoch 00688: val_loss did not improve\n",
      "0s - loss: 0.2760 - acc: 0.9561 - val_loss: 1.3732 - val_acc: 0.8191\n",
      "Epoch 690/1536\n",
      "Epoch 00689: val_loss did not improve\n",
      "0s - loss: 0.9587 - acc: 0.8735 - val_loss: 0.0662 - val_acc: 0.9801\n",
      "Epoch 691/1536\n",
      "Epoch 00690: val_loss did not improve\n",
      "0s - loss: 0.0209 - acc: 0.9932 - val_loss: 0.0563 - val_acc: 0.9839\n",
      "Epoch 692/1536\n",
      "Epoch 00691: val_loss did not improve\n",
      "0s - loss: 0.0152 - acc: 0.9948 - val_loss: 0.0543 - val_acc: 0.9843\n",
      "Epoch 693/1536\n",
      "Epoch 00692: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9955 - val_loss: 0.0525 - val_acc: 0.9851\n",
      "Epoch 694/1536\n",
      "Epoch 00693: val_loss did not improve\n",
      "0s - loss: 0.0131 - acc: 0.9955 - val_loss: 0.0539 - val_acc: 0.9850\n",
      "Epoch 695/1536\n",
      "Epoch 00694: val_loss did not improve\n",
      "0s - loss: 0.0131 - acc: 0.9958 - val_loss: 0.0549 - val_acc: 0.9841\n",
      "Epoch 696/1536\n",
      "Epoch 00695: val_loss did not improve\n",
      "0s - loss: 0.1920 - acc: 0.9718 - val_loss: 0.9909 - val_acc: 0.8781\n",
      "Epoch 697/1536\n",
      "Epoch 00696: val_loss did not improve\n",
      "0s - loss: 0.5143 - acc: 0.9092 - val_loss: 0.0680 - val_acc: 0.9814\n",
      "Epoch 698/1536\n",
      "Epoch 00697: val_loss did not improve\n",
      "0s - loss: 0.0180 - acc: 0.9940 - val_loss: 0.0548 - val_acc: 0.9836\n",
      "Epoch 699/1536\n",
      "Epoch 00698: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9952 - val_loss: 0.0518 - val_acc: 0.9840\n",
      "Epoch 700/1536\n",
      "Epoch 00699: val_loss did not improve\n",
      "0s - loss: 0.0133 - acc: 0.9954 - val_loss: 0.0752 - val_acc: 0.9780\n",
      "Epoch 701/1536\n",
      "Epoch 00700: val_loss did not improve\n",
      "0s - loss: 0.0406 - acc: 0.9879 - val_loss: 0.7216 - val_acc: 0.9304\n",
      "Epoch 702/1536\n",
      "Epoch 00701: val_loss did not improve\n",
      "0s - loss: 0.9218 - acc: 0.8728 - val_loss: 0.5678 - val_acc: 0.9007\n",
      "Epoch 703/1536\n",
      "Epoch 00702: val_loss did not improve\n",
      "0s - loss: 0.1928 - acc: 0.9624 - val_loss: 0.0645 - val_acc: 0.9811\n",
      "Epoch 704/1536\n",
      "Epoch 00703: val_loss did not improve\n",
      "0s - loss: 0.0185 - acc: 0.9937 - val_loss: 0.0610 - val_acc: 0.9824\n",
      "Epoch 705/1536\n",
      "Epoch 00704: val_loss did not improve\n",
      "0s - loss: 0.0140 - acc: 0.9955 - val_loss: 0.0583 - val_acc: 0.9833\n",
      "Epoch 706/1536\n",
      "Epoch 00705: val_loss did not improve\n",
      "0s - loss: 0.0129 - acc: 0.9955 - val_loss: 0.0655 - val_acc: 0.9821\n",
      "Epoch 707/1536\n",
      "Epoch 00706: val_loss did not improve\n",
      "0s - loss: 0.0148 - acc: 0.9953 - val_loss: 0.0659 - val_acc: 0.9826\n",
      "Epoch 708/1536\n",
      "Epoch 00707: val_loss did not improve\n",
      "0s - loss: 0.2576 - acc: 0.9623 - val_loss: 0.3811 - val_acc: 0.9394\n",
      "Epoch 709/1536\n",
      "Epoch 00708: val_loss did not improve\n",
      "0s - loss: 0.2924 - acc: 0.9428 - val_loss: 0.1972 - val_acc: 0.9491\n",
      "Epoch 710/1536\n",
      "Epoch 00709: val_loss did not improve\n",
      "0s - loss: 0.0531 - acc: 0.9833 - val_loss: 0.0554 - val_acc: 0.9835\n",
      "Epoch 711/1536\n",
      "Epoch 00710: val_loss did not improve\n",
      "0s - loss: 0.0116 - acc: 0.9964 - val_loss: 0.0556 - val_acc: 0.9841\n",
      "Epoch 712/1536\n",
      "Epoch 00711: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9970 - val_loss: 0.0591 - val_acc: 0.9824\n",
      "Epoch 713/1536\n",
      "Epoch 00712: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9956 - val_loss: 0.0723 - val_acc: 0.9798\n",
      "Epoch 714/1536\n",
      "Epoch 00713: val_loss did not improve\n",
      "0s - loss: 0.3255 - acc: 0.9401 - val_loss: 0.7071 - val_acc: 0.8594\n",
      "Epoch 715/1536\n",
      "Epoch 00714: val_loss did not improve\n",
      "0s - loss: 1.1188 - acc: 0.8263 - val_loss: 0.0927 - val_acc: 0.9746\n",
      "Epoch 716/1536\n",
      "Epoch 00715: val_loss did not improve\n",
      "0s - loss: 0.0337 - acc: 0.9894 - val_loss: 0.0601 - val_acc: 0.9824\n",
      "Epoch 717/1536\n",
      "Epoch 00716: val_loss did not improve\n",
      "0s - loss: 0.0191 - acc: 0.9933 - val_loss: 0.0550 - val_acc: 0.9842\n",
      "Epoch 718/1536\n",
      "Epoch 00717: val_loss did not improve\n",
      "0s - loss: 0.0168 - acc: 0.9946 - val_loss: 0.0526 - val_acc: 0.9847\n",
      "Epoch 719/1536\n",
      "Epoch 00718: val_loss did not improve\n",
      "0s - loss: 0.0156 - acc: 0.9944 - val_loss: 0.0518 - val_acc: 0.9851\n",
      "Epoch 720/1536\n",
      "Epoch 00719: val_loss did not improve\n",
      "0s - loss: 0.0137 - acc: 0.9953 - val_loss: 0.0564 - val_acc: 0.9842\n",
      "Epoch 721/1536\n",
      "Epoch 00720: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9956 - val_loss: 0.0539 - val_acc: 0.9845\n",
      "Epoch 722/1536\n",
      "Epoch 00721: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9944 - val_loss: 0.1690 - val_acc: 0.9558\n",
      "Epoch 723/1536\n",
      "Epoch 00722: val_loss did not improve\n",
      "0s - loss: 0.4933 - acc: 0.9151 - val_loss: 0.4024 - val_acc: 0.9057\n",
      "Epoch 724/1536\n",
      "Epoch 00723: val_loss did not improve\n",
      "0s - loss: 0.2097 - acc: 0.9570 - val_loss: 0.0602 - val_acc: 0.9817\n",
      "Epoch 725/1536\n",
      "Epoch 00724: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9945 - val_loss: 0.0547 - val_acc: 0.9829\n",
      "Epoch 726/1536\n",
      "Epoch 00725: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9946 - val_loss: 0.0555 - val_acc: 0.9827\n",
      "Epoch 727/1536\n",
      "Epoch 00726: val_loss did not improve\n",
      "0s - loss: 0.0142 - acc: 0.9954 - val_loss: 0.0611 - val_acc: 0.9813\n",
      "Epoch 728/1536\n",
      "Epoch 00727: val_loss improved from 0.05169 to 0.05155, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0179 - acc: 0.9941 - val_loss: 0.0515 - val_acc: 0.9835\n",
      "Epoch 729/1536\n",
      "Epoch 00728: val_loss did not improve\n",
      "0s - loss: 0.0966 - acc: 0.9828 - val_loss: 1.1219 - val_acc: 0.8353\n",
      "Epoch 730/1536\n",
      "Epoch 00729: val_loss did not improve\n",
      "0s - loss: 1.2316 - acc: 0.8267 - val_loss: 0.1047 - val_acc: 0.9731\n",
      "Epoch 731/1536\n",
      "Epoch 00730: val_loss did not improve\n",
      "0s - loss: 0.0282 - acc: 0.9911 - val_loss: 0.0558 - val_acc: 0.9826\n",
      "Epoch 732/1536\n",
      "Epoch 00731: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9944 - val_loss: 0.0549 - val_acc: 0.9833\n",
      "Epoch 733/1536\n",
      "Epoch 00732: val_loss did not improve\n",
      "0s - loss: 0.0154 - acc: 0.9944 - val_loss: 0.0521 - val_acc: 0.9843\n",
      "Epoch 734/1536\n",
      "Epoch 00733: val_loss did not improve\n",
      "0s - loss: 0.0140 - acc: 0.9954 - val_loss: 0.0523 - val_acc: 0.9838\n",
      "Epoch 735/1536\n",
      "Epoch 00734: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9956 - val_loss: 0.0607 - val_acc: 0.9818\n",
      "Epoch 736/1536\n",
      "Epoch 00735: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9959 - val_loss: 0.0596 - val_acc: 0.9821\n",
      "Epoch 737/1536\n",
      "Epoch 00736: val_loss did not improve\n",
      "0s - loss: 0.1177 - acc: 0.9785 - val_loss: 1.1359 - val_acc: 0.8506\n",
      "Epoch 738/1536\n",
      "Epoch 00737: val_loss did not improve\n",
      "0s - loss: 1.4382 - acc: 0.8055 - val_loss: 0.1278 - val_acc: 0.9635\n",
      "Epoch 739/1536\n",
      "Epoch 00738: val_loss did not improve\n",
      "0s - loss: 0.0399 - acc: 0.9866 - val_loss: 0.0626 - val_acc: 0.9816\n",
      "Epoch 740/1536\n",
      "Epoch 00739: val_loss did not improve\n",
      "0s - loss: 0.0241 - acc: 0.9919 - val_loss: 0.0612 - val_acc: 0.9814\n",
      "Epoch 741/1536\n",
      "Epoch 00740: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9935 - val_loss: 0.0579 - val_acc: 0.9823\n",
      "Epoch 742/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00741: val_loss did not improve\n",
      "0s - loss: 0.0186 - acc: 0.9939 - val_loss: 0.0565 - val_acc: 0.9836\n",
      "Epoch 743/1536\n",
      "Epoch 00742: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9947 - val_loss: 0.0568 - val_acc: 0.9831\n",
      "Epoch 744/1536\n",
      "Epoch 00743: val_loss did not improve\n",
      "0s - loss: 0.0271 - acc: 0.9911 - val_loss: 0.1561 - val_acc: 0.9599\n",
      "Epoch 745/1536\n",
      "Epoch 00744: val_loss did not improve\n",
      "0s - loss: 0.6611 - acc: 0.8946 - val_loss: 0.5551 - val_acc: 0.8919\n",
      "Epoch 746/1536\n",
      "Epoch 00745: val_loss improved from 0.05155 to 0.05154, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.1827 - acc: 0.9618 - val_loss: 0.0515 - val_acc: 0.9845\n",
      "Epoch 747/1536\n",
      "Epoch 00746: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9943 - val_loss: 0.0530 - val_acc: 0.9842\n",
      "Epoch 748/1536\n",
      "Epoch 00747: val_loss did not improve\n",
      "0s - loss: 0.0152 - acc: 0.9947 - val_loss: 0.0543 - val_acc: 0.9838\n",
      "Epoch 749/1536\n",
      "Epoch 00748: val_loss did not improve\n",
      "0s - loss: 0.0139 - acc: 0.9955 - val_loss: 0.0586 - val_acc: 0.9837\n",
      "Epoch 750/1536\n",
      "Epoch 00749: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9952 - val_loss: 0.0542 - val_acc: 0.9848\n",
      "Epoch 751/1536\n",
      "Epoch 00750: val_loss did not improve\n",
      "0s - loss: 0.0218 - acc: 0.9930 - val_loss: 0.6915 - val_acc: 0.9141\n",
      "Epoch 752/1536\n",
      "Epoch 00751: val_loss did not improve\n",
      "0s - loss: 0.5397 - acc: 0.9065 - val_loss: 0.3699 - val_acc: 0.9293\n",
      "Epoch 753/1536\n",
      "Epoch 00752: val_loss did not improve\n",
      "0s - loss: 0.1252 - acc: 0.9705 - val_loss: 0.0555 - val_acc: 0.9830\n",
      "Epoch 754/1536\n",
      "Epoch 00753: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9952 - val_loss: 0.0517 - val_acc: 0.9842\n",
      "Epoch 755/1536\n",
      "Epoch 00754: val_loss did not improve\n",
      "0s - loss: 0.0130 - acc: 0.9958 - val_loss: 0.0522 - val_acc: 0.9833\n",
      "Epoch 756/1536\n",
      "Epoch 00755: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9965 - val_loss: 0.0620 - val_acc: 0.9821\n",
      "Epoch 757/1536\n",
      "Epoch 00756: val_loss did not improve\n",
      "0s - loss: 0.1862 - acc: 0.9583 - val_loss: 0.5525 - val_acc: 0.9104\n",
      "Epoch 758/1536\n",
      "Epoch 00757: val_loss did not improve\n",
      "0s - loss: 0.5234 - acc: 0.9065 - val_loss: 0.0675 - val_acc: 0.9806\n",
      "Epoch 759/1536\n",
      "Epoch 00758: val_loss did not improve\n",
      "0s - loss: 0.0204 - acc: 0.9930 - val_loss: 0.0554 - val_acc: 0.9841\n",
      "Epoch 760/1536\n",
      "Epoch 00759: val_loss did not improve\n",
      "0s - loss: 0.0143 - acc: 0.9953 - val_loss: 0.0569 - val_acc: 0.9833\n",
      "Epoch 761/1536\n",
      "Epoch 00760: val_loss did not improve\n",
      "0s - loss: 0.0140 - acc: 0.9954 - val_loss: 0.0574 - val_acc: 0.9831\n",
      "Epoch 762/1536\n",
      "Epoch 00761: val_loss did not improve\n",
      "0s - loss: 0.0153 - acc: 0.9951 - val_loss: 0.0625 - val_acc: 0.9830\n",
      "Epoch 763/1536\n",
      "Epoch 00762: val_loss did not improve\n",
      "0s - loss: 0.0945 - acc: 0.9782 - val_loss: 0.8712 - val_acc: 0.8749\n",
      "Epoch 764/1536\n",
      "Epoch 00763: val_loss did not improve\n",
      "0s - loss: 0.9077 - acc: 0.8571 - val_loss: 0.0834 - val_acc: 0.9768\n",
      "Epoch 765/1536\n",
      "Epoch 00764: val_loss did not improve\n",
      "0s - loss: 0.0300 - acc: 0.9905 - val_loss: 0.0604 - val_acc: 0.9825\n",
      "Epoch 766/1536\n",
      "Epoch 00765: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9949 - val_loss: 0.0629 - val_acc: 0.9821\n",
      "Epoch 767/1536\n",
      "Epoch 00766: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9946 - val_loss: 0.0630 - val_acc: 0.9822\n",
      "Epoch 768/1536\n",
      "Epoch 00767: val_loss did not improve\n",
      "0s - loss: 0.0158 - acc: 0.9951 - val_loss: 0.0551 - val_acc: 0.9843\n",
      "Epoch 769/1536\n",
      "Epoch 00768: val_loss did not improve\n",
      "0s - loss: 0.0143 - acc: 0.9958 - val_loss: 0.0632 - val_acc: 0.9833\n",
      "Epoch 770/1536\n",
      "Epoch 00769: val_loss did not improve\n",
      "0s - loss: 0.1000 - acc: 0.9774 - val_loss: 2.0175 - val_acc: 0.7567\n",
      "Epoch 771/1536\n",
      "Epoch 00770: val_loss did not improve\n",
      "0s - loss: 1.0325 - acc: 0.8535 - val_loss: 0.0633 - val_acc: 0.9815\n",
      "Epoch 772/1536\n",
      "Epoch 00771: val_loss did not improve\n",
      "0s - loss: 0.0175 - acc: 0.9940 - val_loss: 0.0587 - val_acc: 0.9837\n",
      "Epoch 773/1536\n",
      "Epoch 00772: val_loss did not improve\n",
      "0s - loss: 0.0147 - acc: 0.9953 - val_loss: 0.0572 - val_acc: 0.9837\n",
      "Epoch 774/1536\n",
      "Epoch 00773: val_loss did not improve\n",
      "0s - loss: 0.0147 - acc: 0.9953 - val_loss: 0.0576 - val_acc: 0.9836\n",
      "Epoch 775/1536\n",
      "Epoch 00774: val_loss did not improve\n",
      "0s - loss: 0.0129 - acc: 0.9960 - val_loss: 0.0601 - val_acc: 0.9839\n",
      "Epoch 776/1536\n",
      "Epoch 00775: val_loss did not improve\n",
      "0s - loss: 0.0123 - acc: 0.9962 - val_loss: 0.0602 - val_acc: 0.9836\n",
      "Epoch 777/1536\n",
      "Epoch 00776: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9953 - val_loss: 0.0774 - val_acc: 0.9796\n",
      "Epoch 778/1536\n",
      "Epoch 00777: val_loss did not improve\n",
      "0s - loss: 0.4760 - acc: 0.9298 - val_loss: 0.9512 - val_acc: 0.8460\n",
      "Epoch 779/1536\n",
      "Epoch 00778: val_loss did not improve\n",
      "0s - loss: 0.4221 - acc: 0.9234 - val_loss: 0.0653 - val_acc: 0.9822\n",
      "Epoch 780/1536\n",
      "Epoch 00779: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9942 - val_loss: 0.0634 - val_acc: 0.9820\n",
      "Epoch 781/1536\n",
      "Epoch 00780: val_loss did not improve\n",
      "0s - loss: 0.0156 - acc: 0.9951 - val_loss: 0.0632 - val_acc: 0.9828\n",
      "Epoch 782/1536\n",
      "Epoch 00781: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9943 - val_loss: 0.0614 - val_acc: 0.9832\n",
      "Epoch 783/1536\n",
      "Epoch 00782: val_loss did not improve\n",
      "0s - loss: 0.0137 - acc: 0.9956 - val_loss: 0.0588 - val_acc: 0.9822\n",
      "Epoch 784/1536\n",
      "Epoch 00783: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9954 - val_loss: 0.0647 - val_acc: 0.9813\n",
      "Epoch 785/1536\n",
      "Epoch 00784: val_loss did not improve\n",
      "0s - loss: 0.2240 - acc: 0.9675 - val_loss: 0.8803 - val_acc: 0.8692\n",
      "Epoch 786/1536\n",
      "Epoch 00785: val_loss did not improve\n",
      "0s - loss: 0.6581 - acc: 0.8926 - val_loss: 0.0616 - val_acc: 0.9817\n",
      "Epoch 787/1536\n",
      "Epoch 00786: val_loss did not improve\n",
      "0s - loss: 0.0196 - acc: 0.9935 - val_loss: 0.0576 - val_acc: 0.9836\n",
      "Epoch 788/1536\n",
      "Epoch 00787: val_loss did not improve\n",
      "0s - loss: 0.0174 - acc: 0.9937 - val_loss: 0.0603 - val_acc: 0.9824\n",
      "Epoch 789/1536\n",
      "Epoch 00788: val_loss did not improve\n",
      "0s - loss: 0.0148 - acc: 0.9949 - val_loss: 0.0685 - val_acc: 0.9808\n",
      "Epoch 790/1536\n",
      "Epoch 00789: val_loss did not improve\n",
      "0s - loss: 0.0159 - acc: 0.9944 - val_loss: 0.0771 - val_acc: 0.9794\n",
      "Epoch 791/1536\n",
      "Epoch 00790: val_loss did not improve\n",
      "0s - loss: 0.7298 - acc: 0.9061 - val_loss: 0.8162 - val_acc: 0.8623\n",
      "Epoch 792/1536\n",
      "Epoch 00791: val_loss did not improve\n",
      "0s - loss: 0.3990 - acc: 0.9280 - val_loss: 0.0670 - val_acc: 0.9816\n",
      "Epoch 793/1536\n",
      "Epoch 00792: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9945 - val_loss: 0.0606 - val_acc: 0.9835\n",
      "Epoch 794/1536\n",
      "Epoch 00793: val_loss did not improve\n",
      "0s - loss: 0.0161 - acc: 0.9944 - val_loss: 0.0575 - val_acc: 0.9832\n",
      "Epoch 795/1536\n",
      "Epoch 00794: val_loss did not improve\n",
      "0s - loss: 0.0138 - acc: 0.9951 - val_loss: 0.0561 - val_acc: 0.9840\n",
      "Epoch 796/1536\n",
      "Epoch 00795: val_loss did not improve\n",
      "0s - loss: 0.0116 - acc: 0.9962 - val_loss: 0.0580 - val_acc: 0.9838\n",
      "Epoch 797/1536\n",
      "Epoch 00796: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9967 - val_loss: 0.0626 - val_acc: 0.9821\n",
      "Epoch 798/1536\n",
      "Epoch 00797: val_loss did not improve\n",
      "0s - loss: 0.4658 - acc: 0.9373 - val_loss: 0.8098 - val_acc: 0.8627\n",
      "Epoch 799/1536\n",
      "Epoch 00798: val_loss did not improve\n",
      "0s - loss: 0.6991 - acc: 0.8990 - val_loss: 0.0606 - val_acc: 0.9833\n",
      "Epoch 800/1536\n",
      "Epoch 00799: val_loss did not improve\n",
      "0s - loss: 0.0154 - acc: 0.9949 - val_loss: 0.0556 - val_acc: 0.9853\n",
      "Epoch 801/1536\n",
      "Epoch 00800: val_loss did not improve\n",
      "0s - loss: 0.0141 - acc: 0.9956 - val_loss: 0.0535 - val_acc: 0.9846\n",
      "Epoch 802/1536\n",
      "Epoch 00801: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9958 - val_loss: 0.0529 - val_acc: 0.9847\n",
      "Epoch 803/1536\n",
      "Epoch 00802: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9962 - val_loss: 0.0547 - val_acc: 0.9845\n",
      "Epoch 804/1536\n",
      "Epoch 00803: val_loss did not improve\n",
      "0s - loss: 0.0182 - acc: 0.9931 - val_loss: 0.3748 - val_acc: 0.9402\n",
      "Epoch 805/1536\n",
      "Epoch 00804: val_loss did not improve\n",
      "0s - loss: 0.6066 - acc: 0.9032 - val_loss: 0.2832 - val_acc: 0.9351\n",
      "Epoch 806/1536\n",
      "Epoch 00805: val_loss did not improve\n",
      "0s - loss: 0.0635 - acc: 0.9824 - val_loss: 0.0548 - val_acc: 0.9842\n",
      "Epoch 807/1536\n",
      "Epoch 00806: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9954 - val_loss: 0.0576 - val_acc: 0.9847\n",
      "Epoch 808/1536\n",
      "Epoch 00807: val_loss did not improve\n",
      "0s - loss: 0.0133 - acc: 0.9953 - val_loss: 0.0555 - val_acc: 0.9845\n",
      "Epoch 809/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00808: val_loss did not improve\n",
      "0s - loss: 0.0119 - acc: 0.9962 - val_loss: 0.0596 - val_acc: 0.9838\n",
      "Epoch 810/1536\n",
      "Epoch 00809: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9966 - val_loss: 0.0735 - val_acc: 0.9799\n",
      "Epoch 811/1536\n",
      "Epoch 00810: val_loss did not improve\n",
      "0s - loss: 0.9892 - acc: 0.8834 - val_loss: 1.1439 - val_acc: 0.8454\n",
      "Epoch 812/1536\n",
      "Epoch 00811: val_loss did not improve\n",
      "0s - loss: 0.7109 - acc: 0.8917 - val_loss: 0.0901 - val_acc: 0.9768\n",
      "Epoch 813/1536\n",
      "Epoch 00812: val_loss did not improve\n",
      "0s - loss: 0.0268 - acc: 0.9915 - val_loss: 0.0720 - val_acc: 0.9805\n",
      "Epoch 814/1536\n",
      "Epoch 00813: val_loss did not improve\n",
      "0s - loss: 0.0210 - acc: 0.9929 - val_loss: 0.0660 - val_acc: 0.9824\n",
      "Epoch 815/1536\n",
      "Epoch 00814: val_loss did not improve\n",
      "0s - loss: 0.0178 - acc: 0.9945 - val_loss: 0.0653 - val_acc: 0.9815\n",
      "Epoch 816/1536\n",
      "Epoch 00815: val_loss did not improve\n",
      "0s - loss: 0.0136 - acc: 0.9951 - val_loss: 0.0638 - val_acc: 0.9826\n",
      "Epoch 817/1536\n",
      "Epoch 00816: val_loss did not improve\n",
      "0s - loss: 0.0130 - acc: 0.9956 - val_loss: 0.0659 - val_acc: 0.9821\n",
      "Epoch 818/1536\n",
      "Epoch 00817: val_loss did not improve\n",
      "0s - loss: 0.0138 - acc: 0.9952 - val_loss: 0.0606 - val_acc: 0.9829\n",
      "Epoch 819/1536\n",
      "Epoch 00818: val_loss did not improve\n",
      "0s - loss: 0.0187 - acc: 0.9932 - val_loss: 0.4288 - val_acc: 0.9197\n",
      "Epoch 820/1536\n",
      "Epoch 00819: val_loss did not improve\n",
      "0s - loss: 1.2279 - acc: 0.8412 - val_loss: 0.2295 - val_acc: 0.9491\n",
      "Epoch 821/1536\n",
      "Epoch 00820: val_loss did not improve\n",
      "0s - loss: 0.0537 - acc: 0.9850 - val_loss: 0.0622 - val_acc: 0.9824\n",
      "Epoch 822/1536\n",
      "Epoch 00821: val_loss did not improve\n",
      "0s - loss: 0.0152 - acc: 0.9949 - val_loss: 0.0602 - val_acc: 0.9827\n",
      "Epoch 823/1536\n",
      "Epoch 00822: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9953 - val_loss: 0.0580 - val_acc: 0.9825\n",
      "Epoch 824/1536\n",
      "Epoch 00823: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9959 - val_loss: 0.0552 - val_acc: 0.9846\n",
      "Epoch 825/1536\n",
      "Epoch 00824: val_loss did not improve\n",
      "0s - loss: 0.0123 - acc: 0.9957 - val_loss: 0.0709 - val_acc: 0.9793\n",
      "Epoch 826/1536\n",
      "Epoch 00825: val_loss did not improve\n",
      "0s - loss: 0.1312 - acc: 0.9745 - val_loss: 0.4598 - val_acc: 0.9063\n",
      "Epoch 827/1536\n",
      "Epoch 00826: val_loss did not improve\n",
      "0s - loss: 0.6127 - acc: 0.8995 - val_loss: 0.4364 - val_acc: 0.9211\n",
      "Epoch 828/1536\n",
      "Epoch 00827: val_loss did not improve\n",
      "0s - loss: 0.0945 - acc: 0.9787 - val_loss: 0.0570 - val_acc: 0.9839\n",
      "Epoch 829/1536\n",
      "Epoch 00828: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9958 - val_loss: 0.0581 - val_acc: 0.9839\n",
      "Epoch 830/1536\n",
      "Epoch 00829: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9962 - val_loss: 0.0535 - val_acc: 0.9841\n",
      "Epoch 831/1536\n",
      "Epoch 00830: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9962 - val_loss: 0.0539 - val_acc: 0.9859\n",
      "Epoch 832/1536\n",
      "Epoch 00831: val_loss did not improve\n",
      "0s - loss: 0.0094 - acc: 0.9973 - val_loss: 0.0644 - val_acc: 0.9826\n",
      "Epoch 833/1536\n",
      "Epoch 00832: val_loss did not improve\n",
      "0s - loss: 0.1889 - acc: 0.9643 - val_loss: 0.6503 - val_acc: 0.8943\n",
      "Epoch 834/1536\n",
      "Epoch 00833: val_loss did not improve\n",
      "0s - loss: 0.6611 - acc: 0.8918 - val_loss: 0.0624 - val_acc: 0.9829\n",
      "Epoch 835/1536\n",
      "Epoch 00834: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9945 - val_loss: 0.0566 - val_acc: 0.9844\n",
      "Epoch 836/1536\n",
      "Epoch 00835: val_loss did not improve\n",
      "0s - loss: 0.0142 - acc: 0.9955 - val_loss: 0.0547 - val_acc: 0.9859\n",
      "Epoch 837/1536\n",
      "Epoch 00836: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9956 - val_loss: 0.0561 - val_acc: 0.9849\n",
      "Epoch 838/1536\n",
      "Epoch 00837: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9962 - val_loss: 0.0623 - val_acc: 0.9836\n",
      "Epoch 839/1536\n",
      "Epoch 00838: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9948 - val_loss: 0.0688 - val_acc: 0.9807\n",
      "Epoch 840/1536\n",
      "Epoch 00839: val_loss did not improve\n",
      "0s - loss: 0.6829 - acc: 0.9077 - val_loss: 2.4111 - val_acc: 0.7042\n",
      "Epoch 841/1536\n",
      "Epoch 00840: val_loss did not improve\n",
      "0s - loss: 1.2678 - acc: 0.8424 - val_loss: 0.0803 - val_acc: 0.9799\n",
      "Epoch 842/1536\n",
      "Epoch 00841: val_loss did not improve\n",
      "0s - loss: 0.0278 - acc: 0.9915 - val_loss: 0.0688 - val_acc: 0.9829\n",
      "Epoch 843/1536\n",
      "Epoch 00842: val_loss did not improve\n",
      "0s - loss: 0.0201 - acc: 0.9935 - val_loss: 0.0647 - val_acc: 0.9839\n",
      "Epoch 844/1536\n",
      "Epoch 00843: val_loss did not improve\n",
      "0s - loss: 0.0183 - acc: 0.9943 - val_loss: 0.0629 - val_acc: 0.9835\n",
      "Epoch 845/1536\n",
      "Epoch 00844: val_loss did not improve\n",
      "0s - loss: 0.0168 - acc: 0.9944 - val_loss: 0.0707 - val_acc: 0.9815\n",
      "Epoch 846/1536\n",
      "Epoch 00845: val_loss did not improve\n",
      "0s - loss: 0.0158 - acc: 0.9947 - val_loss: 0.0596 - val_acc: 0.9843\n",
      "Epoch 847/1536\n",
      "Epoch 00846: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9959 - val_loss: 0.0592 - val_acc: 0.9831\n",
      "Epoch 848/1536\n",
      "Epoch 00847: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9957 - val_loss: 0.0674 - val_acc: 0.9820\n",
      "Epoch 849/1536\n",
      "Epoch 00848: val_loss did not improve\n",
      "0s - loss: 0.8154 - acc: 0.9005 - val_loss: 0.3303 - val_acc: 0.9380\n",
      "Epoch 850/1536\n",
      "Epoch 00849: val_loss did not improve\n",
      "0s - loss: 0.0996 - acc: 0.9752 - val_loss: 0.0683 - val_acc: 0.9825\n",
      "Epoch 851/1536\n",
      "Epoch 00850: val_loss did not improve\n",
      "0s - loss: 0.0176 - acc: 0.9942 - val_loss: 0.0653 - val_acc: 0.9832\n",
      "Epoch 852/1536\n",
      "Epoch 00851: val_loss did not improve\n",
      "0s - loss: 0.0153 - acc: 0.9946 - val_loss: 0.0689 - val_acc: 0.9820\n",
      "Epoch 853/1536\n",
      "Epoch 00852: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9957 - val_loss: 0.0847 - val_acc: 0.9777\n",
      "Epoch 854/1536\n",
      "Epoch 00853: val_loss did not improve\n",
      "0s - loss: 0.1919 - acc: 0.9588 - val_loss: 0.7973 - val_acc: 0.8940\n",
      "Epoch 855/1536\n",
      "Epoch 00854: val_loss did not improve\n",
      "0s - loss: 0.3328 - acc: 0.9359 - val_loss: 0.0607 - val_acc: 0.9828\n",
      "Epoch 856/1536\n",
      "Epoch 00855: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9959 - val_loss: 0.0580 - val_acc: 0.9850\n",
      "Epoch 857/1536\n",
      "Epoch 00856: val_loss did not improve\n",
      "0s - loss: 0.0112 - acc: 0.9961 - val_loss: 0.0622 - val_acc: 0.9824\n",
      "Epoch 858/1536\n",
      "Epoch 00857: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9961 - val_loss: 0.0609 - val_acc: 0.9829\n",
      "Epoch 859/1536\n",
      "Epoch 00858: val_loss did not improve\n",
      "0s - loss: 0.0785 - acc: 0.9846 - val_loss: 1.1183 - val_acc: 0.8627\n",
      "Epoch 860/1536\n",
      "Epoch 00859: val_loss did not improve\n",
      "0s - loss: 0.6088 - acc: 0.9074 - val_loss: 0.0660 - val_acc: 0.9812\n",
      "Epoch 861/1536\n",
      "Epoch 00860: val_loss did not improve\n",
      "0s - loss: 0.0142 - acc: 0.9951 - val_loss: 0.0578 - val_acc: 0.9834\n",
      "Epoch 862/1536\n",
      "Epoch 00861: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9960 - val_loss: 0.0560 - val_acc: 0.9838\n",
      "Epoch 863/1536\n",
      "Epoch 00862: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9967 - val_loss: 0.0584 - val_acc: 0.9838\n",
      "Epoch 864/1536\n",
      "Epoch 00863: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9965 - val_loss: 0.0597 - val_acc: 0.9828\n",
      "Epoch 865/1536\n",
      "Epoch 00864: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9963 - val_loss: 0.0693 - val_acc: 0.9817\n",
      "Epoch 866/1536\n",
      "Epoch 00865: val_loss did not improve\n",
      "0s - loss: 0.6155 - acc: 0.9242 - val_loss: 1.7836 - val_acc: 0.7932\n",
      "Epoch 867/1536\n",
      "Epoch 00866: val_loss did not improve\n",
      "0s - loss: 0.9615 - acc: 0.8735 - val_loss: 0.0677 - val_acc: 0.9812\n",
      "Epoch 868/1536\n",
      "Epoch 00867: val_loss did not improve\n",
      "0s - loss: 0.0198 - acc: 0.9934 - val_loss: 0.0596 - val_acc: 0.9833\n",
      "Epoch 869/1536\n",
      "Epoch 00868: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9954 - val_loss: 0.0580 - val_acc: 0.9841\n",
      "Epoch 870/1536\n",
      "Epoch 00869: val_loss did not improve\n",
      "0s - loss: 0.0145 - acc: 0.9954 - val_loss: 0.0591 - val_acc: 0.9833\n",
      "Epoch 871/1536\n",
      "Epoch 00870: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9956 - val_loss: 0.0653 - val_acc: 0.9826\n",
      "Epoch 872/1536\n",
      "Epoch 00871: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9954 - val_loss: 0.0739 - val_acc: 0.9812\n",
      "Epoch 873/1536\n",
      "Epoch 00872: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9949 - val_loss: 0.0671 - val_acc: 0.9812\n",
      "Epoch 874/1536\n",
      "Epoch 00873: val_loss did not improve\n",
      "0s - loss: 0.3146 - acc: 0.9559 - val_loss: 1.6535 - val_acc: 0.7917\n",
      "Epoch 875/1536\n",
      "Epoch 00874: val_loss did not improve\n",
      "0s - loss: 0.7126 - acc: 0.8939 - val_loss: 0.0617 - val_acc: 0.9832\n",
      "Epoch 876/1536\n",
      "Epoch 00875: val_loss did not improve\n",
      "0s - loss: 0.0168 - acc: 0.9943 - val_loss: 0.0607 - val_acc: 0.9842\n",
      "Epoch 877/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00876: val_loss did not improve\n",
      "0s - loss: 0.0148 - acc: 0.9949 - val_loss: 0.0565 - val_acc: 0.9851\n",
      "Epoch 878/1536\n",
      "Epoch 00877: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9952 - val_loss: 0.0560 - val_acc: 0.9853\n",
      "Epoch 879/1536\n",
      "Epoch 00878: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9957 - val_loss: 0.0590 - val_acc: 0.9842\n",
      "Epoch 880/1536\n",
      "Epoch 00879: val_loss did not improve\n",
      "0s - loss: 0.0123 - acc: 0.9957 - val_loss: 0.0633 - val_acc: 0.9832\n",
      "Epoch 881/1536\n",
      "Epoch 00880: val_loss did not improve\n",
      "0s - loss: 0.1555 - acc: 0.9742 - val_loss: 0.9749 - val_acc: 0.8516\n",
      "Epoch 882/1536\n",
      "Epoch 00881: val_loss did not improve\n",
      "0s - loss: 0.4598 - acc: 0.9211 - val_loss: 0.1505 - val_acc: 0.9630\n",
      "Epoch 883/1536\n",
      "Epoch 00882: val_loss did not improve\n",
      "0s - loss: 0.0321 - acc: 0.9902 - val_loss: 0.0549 - val_acc: 0.9846\n",
      "Epoch 884/1536\n",
      "Epoch 00883: val_loss did not improve\n",
      "0s - loss: 0.0126 - acc: 0.9953 - val_loss: 0.0571 - val_acc: 0.9844\n",
      "Epoch 885/1536\n",
      "Epoch 00884: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9960 - val_loss: 0.0614 - val_acc: 0.9841\n",
      "Epoch 886/1536\n",
      "Epoch 00885: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9965 - val_loss: 0.0572 - val_acc: 0.9845\n",
      "Epoch 887/1536\n",
      "Epoch 00886: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9958 - val_loss: 0.8133 - val_acc: 0.9105\n",
      "Epoch 888/1536\n",
      "Epoch 00887: val_loss did not improve\n",
      "0s - loss: 1.4882 - acc: 0.8374 - val_loss: 0.2993 - val_acc: 0.9373\n",
      "Epoch 889/1536\n",
      "Epoch 00888: val_loss did not improve\n",
      "0s - loss: 0.1262 - acc: 0.9709 - val_loss: 0.0628 - val_acc: 0.9836\n",
      "Epoch 890/1536\n",
      "Epoch 00889: val_loss did not improve\n",
      "0s - loss: 0.0181 - acc: 0.9937 - val_loss: 0.0573 - val_acc: 0.9852\n",
      "Epoch 891/1536\n",
      "Epoch 00890: val_loss did not improve\n",
      "0s - loss: 0.0142 - acc: 0.9956 - val_loss: 0.0587 - val_acc: 0.9847\n",
      "Epoch 892/1536\n",
      "Epoch 00891: val_loss did not improve\n",
      "0s - loss: 0.0143 - acc: 0.9957 - val_loss: 0.0580 - val_acc: 0.9841\n",
      "Epoch 893/1536\n",
      "Epoch 00892: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9957 - val_loss: 0.0569 - val_acc: 0.9843\n",
      "Epoch 894/1536\n",
      "Epoch 00893: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9964 - val_loss: 0.0969 - val_acc: 0.9757\n",
      "Epoch 895/1536\n",
      "Epoch 00894: val_loss did not improve\n",
      "0s - loss: 0.5290 - acc: 0.9248 - val_loss: 0.5049 - val_acc: 0.9245\n",
      "Epoch 896/1536\n",
      "Epoch 00895: val_loss did not improve\n",
      "0s - loss: 0.1749 - acc: 0.9668 - val_loss: 0.0564 - val_acc: 0.9847\n",
      "Epoch 897/1536\n",
      "Epoch 00896: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9960 - val_loss: 0.0529 - val_acc: 0.9857\n",
      "Epoch 898/1536\n",
      "Epoch 00897: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9957 - val_loss: 0.0594 - val_acc: 0.9836\n",
      "Epoch 899/1536\n",
      "Epoch 00898: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9962 - val_loss: 0.0732 - val_acc: 0.9806\n",
      "Epoch 900/1536\n",
      "Epoch 00899: val_loss did not improve\n",
      "0s - loss: 0.2871 - acc: 0.9640 - val_loss: 0.9258 - val_acc: 0.8682\n",
      "Epoch 901/1536\n",
      "Epoch 00900: val_loss did not improve\n",
      "0s - loss: 0.4566 - acc: 0.9240 - val_loss: 0.0620 - val_acc: 0.9828\n",
      "Epoch 902/1536\n",
      "Epoch 00901: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9956 - val_loss: 0.0567 - val_acc: 0.9842\n",
      "Epoch 903/1536\n",
      "Epoch 00902: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9960 - val_loss: 0.0564 - val_acc: 0.9838\n",
      "Epoch 904/1536\n",
      "Epoch 00903: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9964 - val_loss: 0.0616 - val_acc: 0.9836\n",
      "Epoch 905/1536\n",
      "Epoch 00904: val_loss did not improve\n",
      "0s - loss: 0.0161 - acc: 0.9949 - val_loss: 0.6165 - val_acc: 0.9431\n",
      "Epoch 906/1536\n",
      "Epoch 00905: val_loss did not improve\n",
      "0s - loss: 0.1925 - acc: 0.9740 - val_loss: 0.0587 - val_acc: 0.9833\n",
      "Epoch 907/1536\n",
      "Epoch 00906: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9966 - val_loss: 0.0546 - val_acc: 0.9850\n",
      "Epoch 908/1536\n",
      "Epoch 00907: val_loss did not improve\n",
      "0s - loss: 0.4989 - acc: 0.9389 - val_loss: 1.1903 - val_acc: 0.8548\n",
      "Epoch 909/1536\n",
      "Epoch 00908: val_loss did not improve\n",
      "0s - loss: 0.9280 - acc: 0.8697 - val_loss: 0.0813 - val_acc: 0.9777\n",
      "Epoch 910/1536\n",
      "Epoch 00909: val_loss did not improve\n",
      "0s - loss: 0.0188 - acc: 0.9937 - val_loss: 0.0674 - val_acc: 0.9822\n",
      "Epoch 911/1536\n",
      "Epoch 00910: val_loss did not improve\n",
      "0s - loss: 0.0168 - acc: 0.9940 - val_loss: 0.0628 - val_acc: 0.9833\n",
      "Epoch 912/1536\n",
      "Epoch 00911: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9961 - val_loss: 0.0589 - val_acc: 0.9842\n",
      "Epoch 913/1536\n",
      "Epoch 00912: val_loss did not improve\n",
      "0s - loss: 0.0123 - acc: 0.9960 - val_loss: 0.0571 - val_acc: 0.9847\n",
      "Epoch 914/1536\n",
      "Epoch 00913: val_loss did not improve\n",
      "0s - loss: 0.0115 - acc: 0.9960 - val_loss: 0.0824 - val_acc: 0.9784\n",
      "Epoch 915/1536\n",
      "Epoch 00914: val_loss did not improve\n",
      "0s - loss: 0.2001 - acc: 0.9625 - val_loss: 0.2219 - val_acc: 0.9475\n",
      "Epoch 916/1536\n",
      "Epoch 00915: val_loss did not improve\n",
      "0s - loss: 0.2047 - acc: 0.9530 - val_loss: 0.0596 - val_acc: 0.9837\n",
      "Epoch 917/1536\n",
      "Epoch 00916: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9960 - val_loss: 0.0555 - val_acc: 0.9837\n",
      "Epoch 918/1536\n",
      "Epoch 00917: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9970 - val_loss: 0.0593 - val_acc: 0.9840\n",
      "Epoch 919/1536\n",
      "Epoch 00918: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9973 - val_loss: 0.0578 - val_acc: 0.9840\n",
      "Epoch 920/1536\n",
      "Epoch 00919: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9947 - val_loss: 0.3936 - val_acc: 0.9329\n",
      "Epoch 921/1536\n",
      "Epoch 00920: val_loss did not improve\n",
      "0s - loss: 1.0882 - acc: 0.8563 - val_loss: 0.2865 - val_acc: 0.9409\n",
      "Epoch 922/1536\n",
      "Epoch 00921: val_loss did not improve\n",
      "0s - loss: 0.0620 - acc: 0.9833 - val_loss: 0.0607 - val_acc: 0.9825\n",
      "Epoch 923/1536\n",
      "Epoch 00922: val_loss did not improve\n",
      "0s - loss: 0.0172 - acc: 0.9943 - val_loss: 0.0632 - val_acc: 0.9805\n",
      "Epoch 924/1536\n",
      "Epoch 00923: val_loss did not improve\n",
      "0s - loss: 0.0138 - acc: 0.9951 - val_loss: 0.0611 - val_acc: 0.9828\n",
      "Epoch 925/1536\n",
      "Epoch 00924: val_loss did not improve\n",
      "0s - loss: 0.0136 - acc: 0.9958 - val_loss: 0.0608 - val_acc: 0.9818\n",
      "Epoch 926/1536\n",
      "Epoch 00925: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9955 - val_loss: 0.0653 - val_acc: 0.9805\n",
      "Epoch 927/1536\n",
      "Epoch 00926: val_loss did not improve\n",
      "0s - loss: 0.2254 - acc: 0.9667 - val_loss: 0.5536 - val_acc: 0.8948\n",
      "Epoch 928/1536\n",
      "Epoch 00927: val_loss did not improve\n",
      "0s - loss: 0.3340 - acc: 0.9317 - val_loss: 0.0573 - val_acc: 0.9835\n",
      "Epoch 929/1536\n",
      "Epoch 00928: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9962 - val_loss: 0.0560 - val_acc: 0.9838\n",
      "Epoch 930/1536\n",
      "Epoch 00929: val_loss did not improve\n",
      "0s - loss: 0.0112 - acc: 0.9962 - val_loss: 0.0542 - val_acc: 0.9835\n",
      "Epoch 931/1536\n",
      "Epoch 00930: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9963 - val_loss: 0.0630 - val_acc: 0.9831\n",
      "Epoch 932/1536\n",
      "Epoch 00931: val_loss did not improve\n",
      "0s - loss: 0.0508 - acc: 0.9881 - val_loss: 0.5383 - val_acc: 0.9182\n",
      "Epoch 933/1536\n",
      "Epoch 00932: val_loss did not improve\n",
      "0s - loss: 1.2427 - acc: 0.8437 - val_loss: 0.2945 - val_acc: 0.9508\n",
      "Epoch 934/1536\n",
      "Epoch 00933: val_loss did not improve\n",
      "0s - loss: 0.0981 - acc: 0.9791 - val_loss: 0.0687 - val_acc: 0.9805\n",
      "Epoch 935/1536\n",
      "Epoch 00934: val_loss did not improve\n",
      "0s - loss: 0.0223 - acc: 0.9924 - val_loss: 0.0607 - val_acc: 0.9832\n",
      "Epoch 936/1536\n",
      "Epoch 00935: val_loss did not improve\n",
      "0s - loss: 0.0179 - acc: 0.9935 - val_loss: 0.0612 - val_acc: 0.9825\n",
      "Epoch 937/1536\n",
      "Epoch 00936: val_loss did not improve\n",
      "0s - loss: 0.0143 - acc: 0.9951 - val_loss: 0.0641 - val_acc: 0.9811\n",
      "Epoch 938/1536\n",
      "Epoch 00937: val_loss did not improve\n",
      "0s - loss: 0.0158 - acc: 0.9945 - val_loss: 0.0724 - val_acc: 0.9795\n",
      "Epoch 939/1536\n",
      "Epoch 00938: val_loss did not improve\n",
      "0s - loss: 0.1575 - acc: 0.9714 - val_loss: 0.1253 - val_acc: 0.9682\n",
      "Epoch 940/1536\n",
      "Epoch 00939: val_loss did not improve\n",
      "0s - loss: 0.0334 - acc: 0.9885 - val_loss: 0.0538 - val_acc: 0.9830\n",
      "Epoch 941/1536\n",
      "Epoch 00940: val_loss did not improve\n",
      "0s - loss: 0.0098 - acc: 0.9968 - val_loss: 0.0693 - val_acc: 0.9795\n",
      "Epoch 942/1536\n",
      "Epoch 00941: val_loss did not improve\n",
      "0s - loss: 0.3591 - acc: 0.9479 - val_loss: 1.9519 - val_acc: 0.7998\n",
      "Epoch 943/1536\n",
      "Epoch 00942: val_loss did not improve\n",
      "0s - loss: 1.0179 - acc: 0.8755 - val_loss: 0.0708 - val_acc: 0.9812\n",
      "Epoch 944/1536\n",
      "Epoch 00943: val_loss did not improve\n",
      "0s - loss: 0.0235 - acc: 0.9916 - val_loss: 0.0620 - val_acc: 0.9835\n",
      "Epoch 945/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 00944: val_loss did not improve\n",
      "0s - loss: 0.0211 - acc: 0.9931 - val_loss: 0.0590 - val_acc: 0.9848\n",
      "Epoch 946/1536\n",
      "Epoch 00945: val_loss did not improve\n",
      "0s - loss: 0.0175 - acc: 0.9941 - val_loss: 0.0557 - val_acc: 0.9841\n",
      "Epoch 947/1536\n",
      "Epoch 00946: val_loss did not improve\n",
      "0s - loss: 0.0151 - acc: 0.9950 - val_loss: 0.0523 - val_acc: 0.9847\n",
      "Epoch 948/1536\n",
      "Epoch 00947: val_loss did not improve\n",
      "0s - loss: 0.0129 - acc: 0.9952 - val_loss: 0.0638 - val_acc: 0.9816\n",
      "Epoch 949/1536\n",
      "Epoch 00948: val_loss did not improve\n",
      "0s - loss: 0.5176 - acc: 0.9352 - val_loss: 1.2927 - val_acc: 0.8539\n",
      "Epoch 950/1536\n",
      "Epoch 00949: val_loss did not improve\n",
      "0s - loss: 0.4926 - acc: 0.9337 - val_loss: 0.0683 - val_acc: 0.9831\n",
      "Epoch 951/1536\n",
      "Epoch 00950: val_loss did not improve\n",
      "0s - loss: 0.0151 - acc: 0.9949 - val_loss: 0.0609 - val_acc: 0.9846\n",
      "Epoch 952/1536\n",
      "Epoch 00951: val_loss did not improve\n",
      "0s - loss: 0.0137 - acc: 0.9956 - val_loss: 0.0586 - val_acc: 0.9849\n",
      "Epoch 953/1536\n",
      "Epoch 00952: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9963 - val_loss: 0.0562 - val_acc: 0.9847\n",
      "Epoch 954/1536\n",
      "Epoch 00953: val_loss did not improve\n",
      "0s - loss: 0.0116 - acc: 0.9960 - val_loss: 0.0610 - val_acc: 0.9834\n",
      "Epoch 955/1536\n",
      "Epoch 00954: val_loss did not improve\n",
      "0s - loss: 0.0131 - acc: 0.9954 - val_loss: 0.0933 - val_acc: 0.9754\n",
      "Epoch 956/1536\n",
      "Epoch 00955: val_loss did not improve\n",
      "0s - loss: 0.7652 - acc: 0.8929 - val_loss: 0.3853 - val_acc: 0.9261\n",
      "Epoch 957/1536\n",
      "Epoch 00956: val_loss did not improve\n",
      "0s - loss: 0.1607 - acc: 0.9659 - val_loss: 0.0612 - val_acc: 0.9825\n",
      "Epoch 958/1536\n",
      "Epoch 00957: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9961 - val_loss: 0.0584 - val_acc: 0.9843\n",
      "Epoch 959/1536\n",
      "Epoch 00958: val_loss did not improve\n",
      "0s - loss: 0.0109 - acc: 0.9964 - val_loss: 0.0619 - val_acc: 0.9840\n",
      "Epoch 960/1536\n",
      "Epoch 00959: val_loss did not improve\n",
      "0s - loss: 0.0121 - acc: 0.9958 - val_loss: 0.0605 - val_acc: 0.9835\n",
      "Epoch 961/1536\n",
      "Epoch 00960: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9964 - val_loss: 0.0709 - val_acc: 0.9822\n",
      "Epoch 962/1536\n",
      "Epoch 00961: val_loss did not improve\n",
      "0s - loss: 0.0990 - acc: 0.9783 - val_loss: 0.5740 - val_acc: 0.9093\n",
      "Epoch 963/1536\n",
      "Epoch 00962: val_loss did not improve\n",
      "0s - loss: 0.5143 - acc: 0.9185 - val_loss: 0.1777 - val_acc: 0.9620\n",
      "Epoch 964/1536\n",
      "Epoch 00963: val_loss did not improve\n",
      "0s - loss: 0.0337 - acc: 0.9900 - val_loss: 0.0536 - val_acc: 0.9859\n",
      "Epoch 965/1536\n",
      "Epoch 00964: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9962 - val_loss: 0.0568 - val_acc: 0.9855\n",
      "Epoch 966/1536\n",
      "Epoch 00965: val_loss improved from 0.05154 to 0.05138, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0093 - acc: 0.9971 - val_loss: 0.0514 - val_acc: 0.9861\n",
      "Epoch 967/1536\n",
      "Epoch 00966: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9973 - val_loss: 0.0635 - val_acc: 0.9829\n",
      "Epoch 968/1536\n",
      "Epoch 00967: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9964 - val_loss: 0.5643 - val_acc: 0.9348\n",
      "Epoch 969/1536\n",
      "Epoch 00968: val_loss did not improve\n",
      "0s - loss: 1.6310 - acc: 0.8092 - val_loss: 0.6557 - val_acc: 0.8878\n",
      "Epoch 970/1536\n",
      "Epoch 00969: val_loss did not improve\n",
      "0s - loss: 0.1801 - acc: 0.9641 - val_loss: 0.0669 - val_acc: 0.9831\n",
      "Epoch 971/1536\n",
      "Epoch 00970: val_loss did not improve\n",
      "0s - loss: 0.0203 - acc: 0.9937 - val_loss: 0.0612 - val_acc: 0.9841\n",
      "Epoch 972/1536\n",
      "Epoch 00971: val_loss did not improve\n",
      "0s - loss: 0.0158 - acc: 0.9946 - val_loss: 0.0581 - val_acc: 0.9842\n",
      "Epoch 973/1536\n",
      "Epoch 00972: val_loss did not improve\n",
      "0s - loss: 0.0133 - acc: 0.9956 - val_loss: 0.0663 - val_acc: 0.9827\n",
      "Epoch 974/1536\n",
      "Epoch 00973: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9950 - val_loss: 0.0725 - val_acc: 0.9803\n",
      "Epoch 975/1536\n",
      "Epoch 00974: val_loss did not improve\n",
      "0s - loss: 0.0130 - acc: 0.9954 - val_loss: 0.0551 - val_acc: 0.9844\n",
      "Epoch 976/1536\n",
      "Epoch 00975: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9958 - val_loss: 0.0560 - val_acc: 0.9853\n",
      "Epoch 977/1536\n",
      "Epoch 00976: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9962 - val_loss: 0.0933 - val_acc: 0.9762\n",
      "Epoch 978/1536\n",
      "Epoch 00977: val_loss did not improve\n",
      "0s - loss: 1.0290 - acc: 0.8766 - val_loss: 0.4955 - val_acc: 0.9341\n",
      "Epoch 979/1536\n",
      "Epoch 00978: val_loss did not improve\n",
      "0s - loss: 0.1321 - acc: 0.9756 - val_loss: 0.0622 - val_acc: 0.9836\n",
      "Epoch 980/1536\n",
      "Epoch 00979: val_loss did not improve\n",
      "0s - loss: 0.0251 - acc: 0.9918 - val_loss: 0.0573 - val_acc: 0.9836\n",
      "Epoch 981/1536\n",
      "Epoch 00980: val_loss did not improve\n",
      "0s - loss: 0.0172 - acc: 0.9943 - val_loss: 0.0588 - val_acc: 0.9852\n",
      "Epoch 982/1536\n",
      "Epoch 00981: val_loss did not improve\n",
      "0s - loss: 0.0155 - acc: 0.9949 - val_loss: 0.0560 - val_acc: 0.9844\n",
      "Epoch 983/1536\n",
      "Epoch 00982: val_loss did not improve\n",
      "0s - loss: 0.0131 - acc: 0.9954 - val_loss: 0.0639 - val_acc: 0.9826\n",
      "Epoch 984/1536\n",
      "Epoch 00983: val_loss did not improve\n",
      "0s - loss: 0.0842 - acc: 0.9834 - val_loss: 0.9066 - val_acc: 0.8856\n",
      "Epoch 985/1536\n",
      "Epoch 00984: val_loss did not improve\n",
      "0s - loss: 0.7126 - acc: 0.8912 - val_loss: 0.0714 - val_acc: 0.9808\n",
      "Epoch 986/1536\n",
      "Epoch 00985: val_loss did not improve\n",
      "0s - loss: 0.0174 - acc: 0.9939 - val_loss: 0.0601 - val_acc: 0.9839\n",
      "Epoch 987/1536\n",
      "Epoch 00986: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9957 - val_loss: 0.0597 - val_acc: 0.9842\n",
      "Epoch 988/1536\n",
      "Epoch 00987: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9966 - val_loss: 0.0601 - val_acc: 0.9833\n",
      "Epoch 989/1536\n",
      "Epoch 00988: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9962 - val_loss: 0.0838 - val_acc: 0.9795\n",
      "Epoch 990/1536\n",
      "Epoch 00989: val_loss did not improve\n",
      "0s - loss: 0.6301 - acc: 0.9121 - val_loss: 0.4507 - val_acc: 0.9137\n",
      "Epoch 991/1536\n",
      "Epoch 00990: val_loss did not improve\n",
      "0s - loss: 0.1403 - acc: 0.9690 - val_loss: 0.0576 - val_acc: 0.9834\n",
      "Epoch 992/1536\n",
      "Epoch 00991: val_loss did not improve\n",
      "0s - loss: 0.0136 - acc: 0.9956 - val_loss: 0.0559 - val_acc: 0.9843\n",
      "Epoch 993/1536\n",
      "Epoch 00992: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9963 - val_loss: 0.0551 - val_acc: 0.9847\n",
      "Epoch 994/1536\n",
      "Epoch 00993: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9967 - val_loss: 0.0549 - val_acc: 0.9841\n",
      "Epoch 995/1536\n",
      "Epoch 00994: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9969 - val_loss: 0.0646 - val_acc: 0.9824\n",
      "Epoch 996/1536\n",
      "Epoch 00995: val_loss did not improve\n",
      "0s - loss: 0.2971 - acc: 0.9542 - val_loss: 0.3598 - val_acc: 0.9274\n",
      "Epoch 997/1536\n",
      "Epoch 00996: val_loss did not improve\n",
      "0s - loss: 0.3676 - acc: 0.9298 - val_loss: 0.2914 - val_acc: 0.9481\n",
      "Epoch 998/1536\n",
      "Epoch 00997: val_loss did not improve\n",
      "0s - loss: 0.0585 - acc: 0.9868 - val_loss: 0.0533 - val_acc: 0.9851\n",
      "Epoch 999/1536\n",
      "Epoch 00998: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9964 - val_loss: 0.0532 - val_acc: 0.9847\n",
      "Epoch 1000/1536\n",
      "Epoch 00999: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9965 - val_loss: 0.0572 - val_acc: 0.9838\n",
      "Epoch 1001/1536\n",
      "Epoch 01000: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9971 - val_loss: 0.0548 - val_acc: 0.9845\n",
      "Epoch 1002/1536\n",
      "Epoch 01001: val_loss did not improve\n",
      "0s - loss: 0.0097 - acc: 0.9970 - val_loss: 0.0631 - val_acc: 0.9816\n",
      "Epoch 1003/1536\n",
      "Epoch 01002: val_loss did not improve\n",
      "0s - loss: 0.1349 - acc: 0.9724 - val_loss: 0.9699 - val_acc: 0.8395\n",
      "Epoch 1004/1536\n",
      "Epoch 01003: val_loss did not improve\n",
      "0s - loss: 1.3410 - acc: 0.8216 - val_loss: 0.0897 - val_acc: 0.9772\n",
      "Epoch 1005/1536\n",
      "Epoch 01004: val_loss did not improve\n",
      "0s - loss: 0.0277 - acc: 0.9906 - val_loss: 0.0676 - val_acc: 0.9813\n",
      "Epoch 1006/1536\n",
      "Epoch 01005: val_loss did not improve\n",
      "0s - loss: 0.0179 - acc: 0.9942 - val_loss: 0.0627 - val_acc: 0.9813\n",
      "Epoch 1007/1536\n",
      "Epoch 01006: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9946 - val_loss: 0.0608 - val_acc: 0.9819\n",
      "Epoch 1008/1536\n",
      "Epoch 01007: val_loss did not improve\n",
      "0s - loss: 0.0121 - acc: 0.9960 - val_loss: 0.0596 - val_acc: 0.9824\n",
      "Epoch 1009/1536\n",
      "Epoch 01008: val_loss did not improve\n",
      "0s - loss: 0.0111 - acc: 0.9961 - val_loss: 0.0712 - val_acc: 0.9801\n",
      "Epoch 1010/1536\n",
      "Epoch 01009: val_loss did not improve\n",
      "0s - loss: 0.0183 - acc: 0.9936 - val_loss: 0.0614 - val_acc: 0.9810\n",
      "Epoch 1011/1536\n",
      "Epoch 01010: val_loss did not improve\n",
      "0s - loss: 0.0468 - acc: 0.9889 - val_loss: 0.3372 - val_acc: 0.9290\n",
      "Epoch 1012/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01011: val_loss did not improve\n",
      "0s - loss: 0.4768 - acc: 0.9078 - val_loss: 0.1724 - val_acc: 0.9587\n",
      "Epoch 1013/1536\n",
      "Epoch 01012: val_loss did not improve\n",
      "0s - loss: 0.0425 - acc: 0.9880 - val_loss: 0.0530 - val_acc: 0.9840\n",
      "Epoch 1014/1536\n",
      "Epoch 01013: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9962 - val_loss: 0.0528 - val_acc: 0.9846\n",
      "Epoch 1015/1536\n",
      "Epoch 01014: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9968 - val_loss: 0.0552 - val_acc: 0.9836\n",
      "Epoch 1016/1536\n",
      "Epoch 01015: val_loss did not improve\n",
      "0s - loss: 0.0111 - acc: 0.9961 - val_loss: 0.0572 - val_acc: 0.9823\n",
      "Epoch 1017/1536\n",
      "Epoch 01016: val_loss did not improve\n",
      "0s - loss: 0.0113 - acc: 0.9965 - val_loss: 0.0647 - val_acc: 0.9829\n",
      "Epoch 1018/1536\n",
      "Epoch 01017: val_loss did not improve\n",
      "0s - loss: 0.9048 - acc: 0.8934 - val_loss: 0.2449 - val_acc: 0.9463\n",
      "Epoch 1019/1536\n",
      "Epoch 01018: val_loss did not improve\n",
      "0s - loss: 0.0543 - acc: 0.9861 - val_loss: 0.0568 - val_acc: 0.9840\n",
      "Epoch 1020/1536\n",
      "Epoch 01019: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9951 - val_loss: 0.0566 - val_acc: 0.9839\n",
      "Epoch 1021/1536\n",
      "Epoch 01020: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9957 - val_loss: 0.0585 - val_acc: 0.9837\n",
      "Epoch 1022/1536\n",
      "Epoch 01021: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9964 - val_loss: 0.0566 - val_acc: 0.9846\n",
      "Epoch 1023/1536\n",
      "Epoch 01022: val_loss did not improve\n",
      "0s - loss: 0.0115 - acc: 0.9960 - val_loss: 0.0583 - val_acc: 0.9838\n",
      "Epoch 1024/1536\n",
      "Epoch 01023: val_loss did not improve\n",
      "0s - loss: 0.0611 - acc: 0.9843 - val_loss: 0.2561 - val_acc: 0.9459\n",
      "Epoch 1025/1536\n",
      "Epoch 01024: val_loss did not improve\n",
      "0s - loss: 0.5905 - acc: 0.9077 - val_loss: 0.3458 - val_acc: 0.9310\n",
      "Epoch 1026/1536\n",
      "Epoch 01025: val_loss did not improve\n",
      "0s - loss: 0.2091 - acc: 0.9589 - val_loss: 0.0560 - val_acc: 0.9840\n",
      "Epoch 1027/1536\n",
      "Epoch 01026: val_loss did not improve\n",
      "0s - loss: 0.0133 - acc: 0.9955 - val_loss: 0.0540 - val_acc: 0.9847\n",
      "Epoch 1028/1536\n",
      "Epoch 01027: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9964 - val_loss: 0.0521 - val_acc: 0.9859\n",
      "Epoch 1029/1536\n",
      "Epoch 01028: val_loss did not improve\n",
      "0s - loss: 0.0102 - acc: 0.9966 - val_loss: 0.0573 - val_acc: 0.9839\n",
      "Epoch 1030/1536\n",
      "Epoch 01029: val_loss did not improve\n",
      "0s - loss: 0.0099 - acc: 0.9966 - val_loss: 0.0533 - val_acc: 0.9848\n",
      "Epoch 1031/1536\n",
      "Epoch 01030: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9968 - val_loss: 0.2778 - val_acc: 0.9450\n",
      "Epoch 1032/1536\n",
      "Epoch 01031: val_loss did not improve\n",
      "0s - loss: 0.7932 - acc: 0.8788 - val_loss: 0.2523 - val_acc: 0.9476\n",
      "Epoch 1033/1536\n",
      "Epoch 01032: val_loss did not improve\n",
      "0s - loss: 0.0532 - acc: 0.9853 - val_loss: 0.0607 - val_acc: 0.9830\n",
      "Epoch 1034/1536\n",
      "Epoch 01033: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9949 - val_loss: 0.0628 - val_acc: 0.9820\n",
      "Epoch 1035/1536\n",
      "Epoch 01034: val_loss did not improve\n",
      "0s - loss: 0.0134 - acc: 0.9956 - val_loss: 0.0560 - val_acc: 0.9844\n",
      "Epoch 1036/1536\n",
      "Epoch 01035: val_loss did not improve\n",
      "0s - loss: 0.0115 - acc: 0.9962 - val_loss: 0.0625 - val_acc: 0.9832\n",
      "Epoch 1037/1536\n",
      "Epoch 01036: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9963 - val_loss: 0.0565 - val_acc: 0.9837\n",
      "Epoch 1038/1536\n",
      "Epoch 01037: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9966 - val_loss: 0.0566 - val_acc: 0.9835\n",
      "Epoch 1039/1536\n",
      "Epoch 01038: val_loss did not improve\n",
      "0s - loss: 0.0524 - acc: 0.9882 - val_loss: 0.7081 - val_acc: 0.9068\n",
      "Epoch 1040/1536\n",
      "Epoch 01039: val_loss did not improve\n",
      "0s - loss: 0.4955 - acc: 0.9165 - val_loss: 0.0683 - val_acc: 0.9808\n",
      "Epoch 1041/1536\n",
      "Epoch 01040: val_loss did not improve\n",
      "0s - loss: 0.0141 - acc: 0.9949 - val_loss: 0.0556 - val_acc: 0.9840\n",
      "Epoch 1042/1536\n",
      "Epoch 01041: val_loss did not improve\n",
      "0s - loss: 0.0112 - acc: 0.9966 - val_loss: 0.0523 - val_acc: 0.9852\n",
      "Epoch 1043/1536\n",
      "Epoch 01042: val_loss did not improve\n",
      "0s - loss: 0.0099 - acc: 0.9968 - val_loss: 0.0578 - val_acc: 0.9839\n",
      "Epoch 1044/1536\n",
      "Epoch 01043: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9962 - val_loss: 0.0586 - val_acc: 0.9818\n",
      "Epoch 1045/1536\n",
      "Epoch 01044: val_loss did not improve\n",
      "0s - loss: 0.0639 - acc: 0.9878 - val_loss: 1.1625 - val_acc: 0.8767\n",
      "Epoch 1046/1536\n",
      "Epoch 01045: val_loss did not improve\n",
      "0s - loss: 1.8663 - acc: 0.7923 - val_loss: 0.1206 - val_acc: 0.9725\n",
      "Epoch 1047/1536\n",
      "Epoch 01046: val_loss did not improve\n",
      "0s - loss: 0.0483 - acc: 0.9877 - val_loss: 0.0655 - val_acc: 0.9837\n",
      "Epoch 1048/1536\n",
      "Epoch 01047: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9934 - val_loss: 0.0614 - val_acc: 0.9846\n",
      "Epoch 1049/1536\n",
      "Epoch 01048: val_loss did not improve\n",
      "0s - loss: 0.0189 - acc: 0.9939 - val_loss: 0.0582 - val_acc: 0.9850\n",
      "Epoch 1050/1536\n",
      "Epoch 01049: val_loss did not improve\n",
      "0s - loss: 0.0165 - acc: 0.9947 - val_loss: 0.0614 - val_acc: 0.9839\n",
      "Epoch 1051/1536\n",
      "Epoch 01050: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9957 - val_loss: 0.0582 - val_acc: 0.9845\n",
      "Epoch 1052/1536\n",
      "Epoch 01051: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9958 - val_loss: 0.0637 - val_acc: 0.9833\n",
      "Epoch 1053/1536\n",
      "Epoch 01052: val_loss did not improve\n",
      "0s - loss: 0.0282 - acc: 0.9908 - val_loss: 0.9280 - val_acc: 0.8751\n",
      "Epoch 1054/1536\n",
      "Epoch 01053: val_loss did not improve\n",
      "0s - loss: 0.7429 - acc: 0.8869 - val_loss: 0.0988 - val_acc: 0.9733\n",
      "Epoch 1055/1536\n",
      "Epoch 01054: val_loss did not improve\n",
      "0s - loss: 0.0206 - acc: 0.9927 - val_loss: 0.0560 - val_acc: 0.9844\n",
      "Epoch 1056/1536\n",
      "Epoch 01055: val_loss did not improve\n",
      "0s - loss: 0.0102 - acc: 0.9964 - val_loss: 0.0533 - val_acc: 0.9853\n",
      "Epoch 1057/1536\n",
      "Epoch 01056: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9971 - val_loss: 0.0553 - val_acc: 0.9848\n",
      "Epoch 1058/1536\n",
      "Epoch 01057: val_loss did not improve\n",
      "0s - loss: 0.0096 - acc: 0.9971 - val_loss: 0.0555 - val_acc: 0.9840\n",
      "Epoch 1059/1536\n",
      "Epoch 01058: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9969 - val_loss: 0.0553 - val_acc: 0.9846\n",
      "Epoch 1060/1536\n",
      "Epoch 01059: val_loss did not improve\n",
      "0s - loss: 0.0113 - acc: 0.9956 - val_loss: 0.3082 - val_acc: 0.9474\n",
      "Epoch 1061/1536\n",
      "Epoch 01060: val_loss did not improve\n",
      "0s - loss: 0.8019 - acc: 0.8942 - val_loss: 0.3158 - val_acc: 0.9393\n",
      "Epoch 1062/1536\n",
      "Epoch 01061: val_loss did not improve\n",
      "0s - loss: 0.0788 - acc: 0.9809 - val_loss: 0.0578 - val_acc: 0.9840\n",
      "Epoch 1063/1536\n",
      "Epoch 01062: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9962 - val_loss: 0.0539 - val_acc: 0.9849\n",
      "Epoch 1064/1536\n",
      "Epoch 01063: val_loss did not improve\n",
      "0s - loss: 0.0109 - acc: 0.9964 - val_loss: 0.0530 - val_acc: 0.9853\n",
      "Epoch 1065/1536\n",
      "Epoch 01064: val_loss improved from 0.05138 to 0.05042, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0099 - acc: 0.9967 - val_loss: 0.0504 - val_acc: 0.9853\n",
      "Epoch 1066/1536\n",
      "Epoch 01065: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9963 - val_loss: 0.1016 - val_acc: 0.9758\n",
      "Epoch 1067/1536\n",
      "Epoch 01066: val_loss did not improve\n",
      "0s - loss: 0.7727 - acc: 0.8958 - val_loss: 0.8157 - val_acc: 0.8744\n",
      "Epoch 1068/1536\n",
      "Epoch 01067: val_loss did not improve\n",
      "0s - loss: 0.2707 - acc: 0.9525 - val_loss: 0.0619 - val_acc: 0.9840\n",
      "Epoch 1069/1536\n",
      "Epoch 01068: val_loss did not improve\n",
      "0s - loss: 0.0187 - acc: 0.9939 - val_loss: 0.0617 - val_acc: 0.9835\n",
      "Epoch 1070/1536\n",
      "Epoch 01069: val_loss did not improve\n",
      "0s - loss: 0.0147 - acc: 0.9951 - val_loss: 0.0599 - val_acc: 0.9840\n",
      "Epoch 1071/1536\n",
      "Epoch 01070: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9959 - val_loss: 0.0592 - val_acc: 0.9844\n",
      "Epoch 1072/1536\n",
      "Epoch 01071: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9967 - val_loss: 0.0547 - val_acc: 0.9857\n",
      "Epoch 1073/1536\n",
      "Epoch 01072: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9961 - val_loss: 0.0593 - val_acc: 0.9831\n",
      "Epoch 1074/1536\n",
      "Epoch 01073: val_loss did not improve\n",
      "0s - loss: 0.0162 - acc: 0.9945 - val_loss: 0.6713 - val_acc: 0.9301\n",
      "Epoch 1075/1536\n",
      "Epoch 01074: val_loss did not improve\n",
      "0s - loss: 1.4961 - acc: 0.8175 - val_loss: 0.7647 - val_acc: 0.9008\n",
      "Epoch 1076/1536\n",
      "Epoch 01075: val_loss did not improve\n",
      "0s - loss: 0.2118 - acc: 0.9659 - val_loss: 0.0659 - val_acc: 0.9833\n",
      "Epoch 1077/1536\n",
      "Epoch 01076: val_loss did not improve\n",
      "0s - loss: 0.0193 - acc: 0.9942 - val_loss: 0.0618 - val_acc: 0.9849\n",
      "Epoch 1078/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01077: val_loss did not improve\n",
      "0s - loss: 0.0167 - acc: 0.9941 - val_loss: 0.0592 - val_acc: 0.9844\n",
      "Epoch 1079/1536\n",
      "Epoch 01078: val_loss did not improve\n",
      "0s - loss: 0.0138 - acc: 0.9953 - val_loss: 0.0580 - val_acc: 0.9852\n",
      "Epoch 1080/1536\n",
      "Epoch 01079: val_loss did not improve\n",
      "0s - loss: 0.0121 - acc: 0.9960 - val_loss: 0.0587 - val_acc: 0.9846\n",
      "Epoch 1081/1536\n",
      "Epoch 01080: val_loss did not improve\n",
      "0s - loss: 0.0121 - acc: 0.9960 - val_loss: 0.0547 - val_acc: 0.9851\n",
      "Epoch 1082/1536\n",
      "Epoch 01081: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9962 - val_loss: 0.0556 - val_acc: 0.9860\n",
      "Epoch 1083/1536\n",
      "Epoch 01082: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9969 - val_loss: 0.0725 - val_acc: 0.9815\n",
      "Epoch 1084/1536\n",
      "Epoch 01083: val_loss did not improve\n",
      "0s - loss: 0.6906 - acc: 0.9104 - val_loss: 1.2536 - val_acc: 0.8383\n",
      "Epoch 1085/1536\n",
      "Epoch 01084: val_loss did not improve\n",
      "0s - loss: 0.4849 - acc: 0.9256 - val_loss: 0.0668 - val_acc: 0.9823\n",
      "Epoch 1086/1536\n",
      "Epoch 01085: val_loss did not improve\n",
      "0s - loss: 0.0138 - acc: 0.9952 - val_loss: 0.0671 - val_acc: 0.9829\n",
      "Epoch 1087/1536\n",
      "Epoch 01086: val_loss did not improve\n",
      "0s - loss: 0.0115 - acc: 0.9960 - val_loss: 0.0611 - val_acc: 0.9840\n",
      "Epoch 1088/1536\n",
      "Epoch 01087: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9964 - val_loss: 0.0612 - val_acc: 0.9845\n",
      "Epoch 1089/1536\n",
      "Epoch 01088: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9963 - val_loss: 0.0576 - val_acc: 0.9845\n",
      "Epoch 1090/1536\n",
      "Epoch 01089: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9966 - val_loss: 0.0608 - val_acc: 0.9839\n",
      "Epoch 1091/1536\n",
      "Epoch 01090: val_loss did not improve\n",
      "0s - loss: 0.0231 - acc: 0.9932 - val_loss: 0.6712 - val_acc: 0.9347\n",
      "Epoch 1092/1536\n",
      "Epoch 01091: val_loss did not improve\n",
      "0s - loss: 0.7808 - acc: 0.8985 - val_loss: 0.6120 - val_acc: 0.9070\n",
      "Epoch 1093/1536\n",
      "Epoch 01092: val_loss did not improve\n",
      "0s - loss: 0.1775 - acc: 0.9683 - val_loss: 0.0587 - val_acc: 0.9845\n",
      "Epoch 1094/1536\n",
      "Epoch 01093: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9959 - val_loss: 0.0552 - val_acc: 0.9858\n",
      "Epoch 1095/1536\n",
      "Epoch 01094: val_loss did not improve\n",
      "0s - loss: 0.0096 - acc: 0.9969 - val_loss: 0.0533 - val_acc: 0.9853\n",
      "Epoch 1096/1536\n",
      "Epoch 01095: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9966 - val_loss: 0.0589 - val_acc: 0.9854\n",
      "Epoch 1097/1536\n",
      "Epoch 01096: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9962 - val_loss: 0.0561 - val_acc: 0.9851\n",
      "Epoch 1098/1536\n",
      "Epoch 01097: val_loss did not improve\n",
      "0s - loss: 0.0079 - acc: 0.9975 - val_loss: 0.0558 - val_acc: 0.9851\n",
      "Epoch 1099/1536\n",
      "Epoch 01098: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9971 - val_loss: 0.1446 - val_acc: 0.9644\n",
      "Epoch 1100/1536\n",
      "Epoch 01099: val_loss did not improve\n",
      "0s - loss: 1.3143 - acc: 0.8544 - val_loss: 0.8610 - val_acc: 0.8723\n",
      "Epoch 1101/1536\n",
      "Epoch 01100: val_loss did not improve\n",
      "0s - loss: 0.2955 - acc: 0.9515 - val_loss: 0.0703 - val_acc: 0.9820\n",
      "Epoch 1102/1536\n",
      "Epoch 01101: val_loss did not improve\n",
      "0s - loss: 0.0228 - acc: 0.9927 - val_loss: 0.0653 - val_acc: 0.9837\n",
      "Epoch 1103/1536\n",
      "Epoch 01102: val_loss did not improve\n",
      "0s - loss: 0.0191 - acc: 0.9933 - val_loss: 0.0616 - val_acc: 0.9848\n",
      "Epoch 1104/1536\n",
      "Epoch 01103: val_loss did not improve\n",
      "0s - loss: 0.0163 - acc: 0.9944 - val_loss: 0.0615 - val_acc: 0.9832\n",
      "Epoch 1105/1536\n",
      "Epoch 01104: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9956 - val_loss: 0.0649 - val_acc: 0.9840\n",
      "Epoch 1106/1536\n",
      "Epoch 01105: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9952 - val_loss: 0.0715 - val_acc: 0.9826\n",
      "Epoch 1107/1536\n",
      "Epoch 01106: val_loss did not improve\n",
      "0s - loss: 0.0207 - acc: 0.9926 - val_loss: 0.5838 - val_acc: 0.9214\n",
      "Epoch 1108/1536\n",
      "Epoch 01107: val_loss did not improve\n",
      "0s - loss: 1.0854 - acc: 0.8646 - val_loss: 0.1644 - val_acc: 0.9640\n",
      "Epoch 1109/1536\n",
      "Epoch 01108: val_loss did not improve\n",
      "0s - loss: 0.0348 - acc: 0.9893 - val_loss: 0.0715 - val_acc: 0.9824\n",
      "Epoch 1110/1536\n",
      "Epoch 01109: val_loss did not improve\n",
      "0s - loss: 0.0136 - acc: 0.9950 - val_loss: 0.0662 - val_acc: 0.9834\n",
      "Epoch 1111/1536\n",
      "Epoch 01110: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9954 - val_loss: 0.0675 - val_acc: 0.9826\n",
      "Epoch 1112/1536\n",
      "Epoch 01111: val_loss did not improve\n",
      "0s - loss: 0.0111 - acc: 0.9963 - val_loss: 0.0660 - val_acc: 0.9841\n",
      "Epoch 1113/1536\n",
      "Epoch 01112: val_loss did not improve\n",
      "0s - loss: 0.0110 - acc: 0.9965 - val_loss: 0.0663 - val_acc: 0.9838\n",
      "Epoch 1114/1536\n",
      "Epoch 01113: val_loss did not improve\n",
      "0s - loss: 0.0113 - acc: 0.9960 - val_loss: 0.0696 - val_acc: 0.9819\n",
      "Epoch 1115/1536\n",
      "Epoch 01114: val_loss did not improve\n",
      "0s - loss: 0.6833 - acc: 0.9247 - val_loss: 1.4202 - val_acc: 0.8278\n",
      "Epoch 1116/1536\n",
      "Epoch 01115: val_loss did not improve\n",
      "0s - loss: 0.5148 - acc: 0.9259 - val_loss: 0.0707 - val_acc: 0.9827\n",
      "Epoch 1117/1536\n",
      "Epoch 01116: val_loss did not improve\n",
      "0s - loss: 0.0119 - acc: 0.9958 - val_loss: 0.0654 - val_acc: 0.9837\n",
      "Epoch 1118/1536\n",
      "Epoch 01117: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9966 - val_loss: 0.0638 - val_acc: 0.9828\n",
      "Epoch 1119/1536\n",
      "Epoch 01118: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9966 - val_loss: 0.0676 - val_acc: 0.9827\n",
      "Epoch 1120/1536\n",
      "Epoch 01119: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9965 - val_loss: 0.0643 - val_acc: 0.9840\n",
      "Epoch 1121/1536\n",
      "Epoch 01120: val_loss did not improve\n",
      "0s - loss: 0.0099 - acc: 0.9968 - val_loss: 0.0667 - val_acc: 0.9830\n",
      "Epoch 1122/1536\n",
      "Epoch 01121: val_loss did not improve\n",
      "0s - loss: 0.2187 - acc: 0.9682 - val_loss: 0.8897 - val_acc: 0.9010\n",
      "Epoch 1123/1536\n",
      "Epoch 01122: val_loss did not improve\n",
      "0s - loss: 1.4076 - acc: 0.8382 - val_loss: 0.0799 - val_acc: 0.9804\n",
      "Epoch 1124/1536\n",
      "Epoch 01123: val_loss did not improve\n",
      "0s - loss: 0.0186 - acc: 0.9939 - val_loss: 0.0656 - val_acc: 0.9829\n",
      "Epoch 1125/1536\n",
      "Epoch 01124: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9956 - val_loss: 0.0648 - val_acc: 0.9841\n",
      "Epoch 1126/1536\n",
      "Epoch 01125: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9955 - val_loss: 0.0599 - val_acc: 0.9841\n",
      "Epoch 1127/1536\n",
      "Epoch 01126: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9966 - val_loss: 0.0584 - val_acc: 0.9845\n",
      "Epoch 1128/1536\n",
      "Epoch 01127: val_loss did not improve\n",
      "0s - loss: 0.0099 - acc: 0.9964 - val_loss: 0.0597 - val_acc: 0.9842\n",
      "Epoch 1129/1536\n",
      "Epoch 01128: val_loss did not improve\n",
      "0s - loss: 0.0096 - acc: 0.9967 - val_loss: 0.0712 - val_acc: 0.9807\n",
      "Epoch 1130/1536\n",
      "Epoch 01129: val_loss did not improve\n",
      "0s - loss: 0.0262 - acc: 0.9931 - val_loss: 0.6359 - val_acc: 0.9380\n",
      "Epoch 1131/1536\n",
      "Epoch 01130: val_loss did not improve\n",
      "0s - loss: 1.1058 - acc: 0.8665 - val_loss: 0.6372 - val_acc: 0.9009\n",
      "Epoch 1132/1536\n",
      "Epoch 01131: val_loss did not improve\n",
      "0s - loss: 0.1709 - acc: 0.9688 - val_loss: 0.0609 - val_acc: 0.9841\n",
      "Epoch 1133/1536\n",
      "Epoch 01132: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9956 - val_loss: 0.0584 - val_acc: 0.9847\n",
      "Epoch 1134/1536\n",
      "Epoch 01133: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9962 - val_loss: 0.0601 - val_acc: 0.9830\n",
      "Epoch 1135/1536\n",
      "Epoch 01134: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9966 - val_loss: 0.0550 - val_acc: 0.9859\n",
      "Epoch 1136/1536\n",
      "Epoch 01135: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9971 - val_loss: 0.0693 - val_acc: 0.9823\n",
      "Epoch 1137/1536\n",
      "Epoch 01136: val_loss did not improve\n",
      "0s - loss: 0.0116 - acc: 0.9960 - val_loss: 0.0695 - val_acc: 0.9824\n",
      "Epoch 1138/1536\n",
      "Epoch 01137: val_loss did not improve\n",
      "0s - loss: 0.5504 - acc: 0.9353 - val_loss: 0.9491 - val_acc: 0.8651\n",
      "Epoch 1139/1536\n",
      "Epoch 01138: val_loss did not improve\n",
      "0s - loss: 0.3377 - acc: 0.9454 - val_loss: 0.0652 - val_acc: 0.9840\n",
      "Epoch 1140/1536\n",
      "Epoch 01139: val_loss did not improve\n",
      "0s - loss: 0.0144 - acc: 0.9957 - val_loss: 0.0621 - val_acc: 0.9840\n",
      "Epoch 1141/1536\n",
      "Epoch 01140: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9963 - val_loss: 0.0664 - val_acc: 0.9825\n",
      "Epoch 1142/1536\n",
      "Epoch 01141: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9961 - val_loss: 0.0742 - val_acc: 0.9812\n",
      "Epoch 1143/1536\n",
      "Epoch 01142: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9957 - val_loss: 0.0659 - val_acc: 0.9828\n",
      "Epoch 1144/1536\n",
      "Epoch 01143: val_loss did not improve\n",
      "0s - loss: 0.0125 - acc: 0.9959 - val_loss: 0.0582 - val_acc: 0.9849\n",
      "Epoch 1145/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01144: val_loss did not improve\n",
      "0s - loss: 0.0632 - acc: 0.9889 - val_loss: 0.8477 - val_acc: 0.9181\n",
      "Epoch 1146/1536\n",
      "Epoch 01145: val_loss did not improve\n",
      "0s - loss: 0.9096 - acc: 0.8851 - val_loss: 0.0774 - val_acc: 0.9814\n",
      "Epoch 1147/1536\n",
      "Epoch 01146: val_loss did not improve\n",
      "0s - loss: 0.0142 - acc: 0.9953 - val_loss: 0.0620 - val_acc: 0.9831\n",
      "Epoch 1148/1536\n",
      "Epoch 01147: val_loss did not improve\n",
      "0s - loss: 0.0135 - acc: 0.9950 - val_loss: 0.0611 - val_acc: 0.9838\n",
      "Epoch 1149/1536\n",
      "Epoch 01148: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9964 - val_loss: 0.0581 - val_acc: 0.9857\n",
      "Epoch 1150/1536\n",
      "Epoch 01149: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9962 - val_loss: 0.0790 - val_acc: 0.9807\n",
      "Epoch 1151/1536\n",
      "Epoch 01150: val_loss did not improve\n",
      "0s - loss: 0.2131 - acc: 0.9684 - val_loss: 0.2906 - val_acc: 0.9418\n",
      "Epoch 1152/1536\n",
      "Epoch 01151: val_loss did not improve\n",
      "0s - loss: 0.4159 - acc: 0.9312 - val_loss: 0.0737 - val_acc: 0.9813\n",
      "Epoch 1153/1536\n",
      "Epoch 01152: val_loss did not improve\n",
      "0s - loss: 0.0125 - acc: 0.9960 - val_loss: 0.0599 - val_acc: 0.9838\n",
      "Epoch 1154/1536\n",
      "Epoch 01153: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9969 - val_loss: 0.0564 - val_acc: 0.9852\n",
      "Epoch 1155/1536\n",
      "Epoch 01154: val_loss did not improve\n",
      "0s - loss: 0.0078 - acc: 0.9970 - val_loss: 0.0568 - val_acc: 0.9851\n",
      "Epoch 1156/1536\n",
      "Epoch 01155: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9969 - val_loss: 0.1082 - val_acc: 0.9754\n",
      "Epoch 1157/1536\n",
      "Epoch 01156: val_loss did not improve\n",
      "0s - loss: 0.4482 - acc: 0.9375 - val_loss: 0.6225 - val_acc: 0.8998\n",
      "Epoch 1158/1536\n",
      "Epoch 01157: val_loss did not improve\n",
      "0s - loss: 0.4532 - acc: 0.9310 - val_loss: 0.0578 - val_acc: 0.9851\n",
      "Epoch 1159/1536\n",
      "Epoch 01158: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9966 - val_loss: 0.0571 - val_acc: 0.9851\n",
      "Epoch 1160/1536\n",
      "Epoch 01159: val_loss did not improve\n",
      "0s - loss: 0.0093 - acc: 0.9968 - val_loss: 0.0527 - val_acc: 0.9862\n",
      "Epoch 1161/1536\n",
      "Epoch 01160: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9970 - val_loss: 0.0532 - val_acc: 0.9863\n",
      "Epoch 1162/1536\n",
      "Epoch 01161: val_loss did not improve\n",
      "0s - loss: 0.0078 - acc: 0.9974 - val_loss: 0.0517 - val_acc: 0.9862\n",
      "Epoch 1163/1536\n",
      "Epoch 01162: val_loss improved from 0.05042 to 0.04949, saving model to saved_models/resnet50_rgbModel_197px_batch8192_epoch1536.hdf5\n",
      "0s - loss: 0.0069 - acc: 0.9979 - val_loss: 0.0495 - val_acc: 0.9876\n",
      "Epoch 1164/1536\n",
      "Epoch 01163: val_loss did not improve\n",
      "0s - loss: 0.0080 - acc: 0.9972 - val_loss: 0.1064 - val_acc: 0.9764\n",
      "Epoch 1165/1536\n",
      "Epoch 01164: val_loss did not improve\n",
      "0s - loss: 0.7459 - acc: 0.9035 - val_loss: 1.0845 - val_acc: 0.8747\n",
      "Epoch 1166/1536\n",
      "Epoch 01165: val_loss did not improve\n",
      "0s - loss: 0.2372 - acc: 0.9650 - val_loss: 0.0623 - val_acc: 0.9841\n",
      "Epoch 1167/1536\n",
      "Epoch 01166: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9956 - val_loss: 0.0591 - val_acc: 0.9843\n",
      "Epoch 1168/1536\n",
      "Epoch 01167: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9960 - val_loss: 0.0652 - val_acc: 0.9822\n",
      "Epoch 1169/1536\n",
      "Epoch 01168: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9966 - val_loss: 0.0560 - val_acc: 0.9852\n",
      "Epoch 1170/1536\n",
      "Epoch 01169: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9971 - val_loss: 0.0563 - val_acc: 0.9841\n",
      "Epoch 1171/1536\n",
      "Epoch 01170: val_loss did not improve\n",
      "0s - loss: 0.0087 - acc: 0.9971 - val_loss: 0.0609 - val_acc: 0.9837\n",
      "Epoch 1172/1536\n",
      "Epoch 01171: val_loss did not improve\n",
      "0s - loss: 0.5248 - acc: 0.9394 - val_loss: 2.5596 - val_acc: 0.7263\n",
      "Epoch 1173/1536\n",
      "Epoch 01172: val_loss did not improve\n",
      "0s - loss: 1.6265 - acc: 0.8174 - val_loss: 0.0763 - val_acc: 0.9813\n",
      "Epoch 1174/1536\n",
      "Epoch 01173: val_loss did not improve\n",
      "0s - loss: 0.0190 - acc: 0.9942 - val_loss: 0.0672 - val_acc: 0.9830\n",
      "Epoch 1175/1536\n",
      "Epoch 01174: val_loss did not improve\n",
      "0s - loss: 0.0145 - acc: 0.9951 - val_loss: 0.0652 - val_acc: 0.9832\n",
      "Epoch 1176/1536\n",
      "Epoch 01175: val_loss did not improve\n",
      "0s - loss: 0.0126 - acc: 0.9959 - val_loss: 0.0623 - val_acc: 0.9842\n",
      "Epoch 1177/1536\n",
      "Epoch 01176: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9963 - val_loss: 0.0648 - val_acc: 0.9837\n",
      "Epoch 1178/1536\n",
      "Epoch 01177: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9963 - val_loss: 0.0643 - val_acc: 0.9832\n",
      "Epoch 1179/1536\n",
      "Epoch 01178: val_loss did not improve\n",
      "0s - loss: 0.0141 - acc: 0.9952 - val_loss: 0.0660 - val_acc: 0.9837\n",
      "Epoch 1180/1536\n",
      "Epoch 01179: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9964 - val_loss: 0.0948 - val_acc: 0.9773\n",
      "Epoch 1181/1536\n",
      "Epoch 01180: val_loss did not improve\n",
      "0s - loss: 0.1170 - acc: 0.9790 - val_loss: 0.4760 - val_acc: 0.9174\n",
      "Epoch 1182/1536\n",
      "Epoch 01181: val_loss did not improve\n",
      "0s - loss: 0.6905 - acc: 0.8934 - val_loss: 0.2284 - val_acc: 0.9554\n",
      "Epoch 1183/1536\n",
      "Epoch 01182: val_loss did not improve\n",
      "0s - loss: 0.0460 - acc: 0.9883 - val_loss: 0.0576 - val_acc: 0.9846\n",
      "Epoch 1184/1536\n",
      "Epoch 01183: val_loss did not improve\n",
      "0s - loss: 0.0090 - acc: 0.9971 - val_loss: 0.0601 - val_acc: 0.9845\n",
      "Epoch 1185/1536\n",
      "Epoch 01184: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9970 - val_loss: 0.0547 - val_acc: 0.9851\n",
      "Epoch 1186/1536\n",
      "Epoch 01185: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9976 - val_loss: 0.0566 - val_acc: 0.9847\n",
      "Epoch 1187/1536\n",
      "Epoch 01186: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9977 - val_loss: 0.0576 - val_acc: 0.9846\n",
      "Epoch 1188/1536\n",
      "Epoch 01187: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9971 - val_loss: 0.0823 - val_acc: 0.9809\n",
      "Epoch 1189/1536\n",
      "Epoch 01188: val_loss did not improve\n",
      "0s - loss: 1.0517 - acc: 0.8860 - val_loss: 0.5598 - val_acc: 0.9089\n",
      "Epoch 1190/1536\n",
      "Epoch 01189: val_loss did not improve\n",
      "0s - loss: 0.1739 - acc: 0.9681 - val_loss: 0.0632 - val_acc: 0.9846\n",
      "Epoch 1191/1536\n",
      "Epoch 01190: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9960 - val_loss: 0.0604 - val_acc: 0.9851\n",
      "Epoch 1192/1536\n",
      "Epoch 01191: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9968 - val_loss: 0.0640 - val_acc: 0.9847\n",
      "Epoch 1193/1536\n",
      "Epoch 01192: val_loss did not improve\n",
      "0s - loss: 0.0112 - acc: 0.9964 - val_loss: 0.0592 - val_acc: 0.9858\n",
      "Epoch 1194/1536\n",
      "Epoch 01193: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9966 - val_loss: 0.0564 - val_acc: 0.9865\n",
      "Epoch 1195/1536\n",
      "Epoch 01194: val_loss did not improve\n",
      "0s - loss: 0.0098 - acc: 0.9965 - val_loss: 0.0595 - val_acc: 0.9847\n",
      "Epoch 1196/1536\n",
      "Epoch 01195: val_loss did not improve\n",
      "0s - loss: 0.0554 - acc: 0.9879 - val_loss: 0.2507 - val_acc: 0.9496\n",
      "Epoch 1197/1536\n",
      "Epoch 01196: val_loss did not improve\n",
      "0s - loss: 0.4935 - acc: 0.9242 - val_loss: 0.2234 - val_acc: 0.9548\n",
      "Epoch 1198/1536\n",
      "Epoch 01197: val_loss did not improve\n",
      "0s - loss: 0.0558 - acc: 0.9862 - val_loss: 0.0579 - val_acc: 0.9853\n",
      "Epoch 1199/1536\n",
      "Epoch 01198: val_loss did not improve\n",
      "0s - loss: 0.0093 - acc: 0.9969 - val_loss: 0.0577 - val_acc: 0.9852\n",
      "Epoch 1200/1536\n",
      "Epoch 01199: val_loss did not improve\n",
      "0s - loss: 0.0075 - acc: 0.9975 - val_loss: 0.0570 - val_acc: 0.9858\n",
      "Epoch 1201/1536\n",
      "Epoch 01200: val_loss did not improve\n",
      "0s - loss: 0.0092 - acc: 0.9971 - val_loss: 0.0618 - val_acc: 0.9841\n",
      "Epoch 1202/1536\n",
      "Epoch 01201: val_loss did not improve\n",
      "0s - loss: 0.0080 - acc: 0.9972 - val_loss: 0.0659 - val_acc: 0.9837\n",
      "Epoch 1203/1536\n",
      "Epoch 01202: val_loss did not improve\n",
      "0s - loss: 0.6448 - acc: 0.9293 - val_loss: 0.8970 - val_acc: 0.9138\n",
      "Epoch 1204/1536\n",
      "Epoch 01203: val_loss did not improve\n",
      "0s - loss: 0.3854 - acc: 0.9501 - val_loss: 0.0673 - val_acc: 0.9832\n",
      "Epoch 1205/1536\n",
      "Epoch 01204: val_loss did not improve\n",
      "0s - loss: 0.0175 - acc: 0.9944 - val_loss: 0.0701 - val_acc: 0.9832\n",
      "Epoch 1206/1536\n",
      "Epoch 01205: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9953 - val_loss: 0.0591 - val_acc: 0.9855\n",
      "Epoch 1207/1536\n",
      "Epoch 01206: val_loss did not improve\n",
      "0s - loss: 0.0125 - acc: 0.9962 - val_loss: 0.0570 - val_acc: 0.9852\n",
      "Epoch 1208/1536\n",
      "Epoch 01207: val_loss did not improve\n",
      "0s - loss: 0.0089 - acc: 0.9971 - val_loss: 0.0609 - val_acc: 0.9847\n",
      "Epoch 1209/1536\n",
      "Epoch 01208: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9971 - val_loss: 0.0671 - val_acc: 0.9838\n",
      "Epoch 1210/1536\n",
      "Epoch 01209: val_loss did not improve\n",
      "0s - loss: 0.0136 - acc: 0.9956 - val_loss: 0.5532 - val_acc: 0.9124\n",
      "Epoch 1211/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01210: val_loss did not improve\n",
      "0s - loss: 1.0787 - acc: 0.8702 - val_loss: 0.0790 - val_acc: 0.9790\n",
      "Epoch 1212/1536\n",
      "Epoch 01211: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9944 - val_loss: 0.0549 - val_acc: 0.9846\n",
      "Epoch 1213/1536\n",
      "Epoch 01212: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9962 - val_loss: 0.0567 - val_acc: 0.9853\n",
      "Epoch 1214/1536\n",
      "Epoch 01213: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9964 - val_loss: 0.0555 - val_acc: 0.9854\n",
      "Epoch 1215/1536\n",
      "Epoch 01214: val_loss did not improve\n",
      "0s - loss: 0.0084 - acc: 0.9972 - val_loss: 0.0596 - val_acc: 0.9849\n",
      "Epoch 1216/1536\n",
      "Epoch 01215: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9971 - val_loss: 0.0772 - val_acc: 0.9818\n",
      "Epoch 1217/1536\n",
      "Epoch 01216: val_loss did not improve\n",
      "0s - loss: 0.1220 - acc: 0.9797 - val_loss: 1.9744 - val_acc: 0.7888\n",
      "Epoch 1218/1536\n",
      "Epoch 01217: val_loss did not improve\n",
      "0s - loss: 1.6494 - acc: 0.8161 - val_loss: 0.0870 - val_acc: 0.9799\n",
      "Epoch 1219/1536\n",
      "Epoch 01218: val_loss did not improve\n",
      "0s - loss: 0.0213 - acc: 0.9934 - val_loss: 0.0691 - val_acc: 0.9834\n",
      "Epoch 1220/1536\n",
      "Epoch 01219: val_loss did not improve\n",
      "0s - loss: 0.0150 - acc: 0.9951 - val_loss: 0.0651 - val_acc: 0.9842\n",
      "Epoch 1221/1536\n",
      "Epoch 01220: val_loss did not improve\n",
      "0s - loss: 0.0125 - acc: 0.9955 - val_loss: 0.0638 - val_acc: 0.9844\n",
      "Epoch 1222/1536\n",
      "Epoch 01221: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9956 - val_loss: 0.0620 - val_acc: 0.9849\n",
      "Epoch 1223/1536\n",
      "Epoch 01222: val_loss did not improve\n",
      "0s - loss: 0.0092 - acc: 0.9971 - val_loss: 0.0703 - val_acc: 0.9831\n",
      "Epoch 1224/1536\n",
      "Epoch 01223: val_loss did not improve\n",
      "0s - loss: 0.0174 - acc: 0.9937 - val_loss: 0.1150 - val_acc: 0.9746\n",
      "Epoch 1225/1536\n",
      "Epoch 01224: val_loss did not improve\n",
      "0s - loss: 0.3830 - acc: 0.9424 - val_loss: 0.2052 - val_acc: 0.9593\n",
      "Epoch 1226/1536\n",
      "Epoch 01225: val_loss did not improve\n",
      "0s - loss: 0.0386 - acc: 0.9892 - val_loss: 0.0604 - val_acc: 0.9844\n",
      "Epoch 1227/1536\n",
      "Epoch 01226: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9974 - val_loss: 0.0615 - val_acc: 0.9850\n",
      "Epoch 1228/1536\n",
      "Epoch 01227: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9973 - val_loss: 0.0623 - val_acc: 0.9840\n",
      "Epoch 1229/1536\n",
      "Epoch 01228: val_loss did not improve\n",
      "0s - loss: 0.0077 - acc: 0.9974 - val_loss: 0.0692 - val_acc: 0.9824\n",
      "Epoch 1230/1536\n",
      "Epoch 01229: val_loss did not improve\n",
      "0s - loss: 0.0117 - acc: 0.9959 - val_loss: 0.6168 - val_acc: 0.9141\n",
      "Epoch 1231/1536\n",
      "Epoch 01230: val_loss did not improve\n",
      "0s - loss: 1.1072 - acc: 0.8690 - val_loss: 0.1559 - val_acc: 0.9672\n",
      "Epoch 1232/1536\n",
      "Epoch 01231: val_loss did not improve\n",
      "0s - loss: 0.0425 - acc: 0.9881 - val_loss: 0.0715 - val_acc: 0.9826\n",
      "Epoch 1233/1536\n",
      "Epoch 01232: val_loss did not improve\n",
      "0s - loss: 0.0177 - acc: 0.9944 - val_loss: 0.0694 - val_acc: 0.9836\n",
      "Epoch 1234/1536\n",
      "Epoch 01233: val_loss did not improve\n",
      "0s - loss: 0.0148 - acc: 0.9950 - val_loss: 0.0613 - val_acc: 0.9852\n",
      "Epoch 1235/1536\n",
      "Epoch 01234: val_loss did not improve\n",
      "0s - loss: 0.0116 - acc: 0.9960 - val_loss: 0.0630 - val_acc: 0.9849\n",
      "Epoch 1236/1536\n",
      "Epoch 01235: val_loss did not improve\n",
      "0s - loss: 0.0110 - acc: 0.9962 - val_loss: 0.0853 - val_acc: 0.9799\n",
      "Epoch 1237/1536\n",
      "Epoch 01236: val_loss did not improve\n",
      "0s - loss: 0.3723 - acc: 0.9421 - val_loss: 0.7489 - val_acc: 0.9092\n",
      "Epoch 1238/1536\n",
      "Epoch 01237: val_loss did not improve\n",
      "0s - loss: 0.2236 - acc: 0.9637 - val_loss: 0.0614 - val_acc: 0.9835\n",
      "Epoch 1239/1536\n",
      "Epoch 01238: val_loss did not improve\n",
      "0s - loss: 0.0081 - acc: 0.9971 - val_loss: 0.0599 - val_acc: 0.9837\n",
      "Epoch 1240/1536\n",
      "Epoch 01239: val_loss did not improve\n",
      "0s - loss: 0.0089 - acc: 0.9970 - val_loss: 0.0638 - val_acc: 0.9830\n",
      "Epoch 1241/1536\n",
      "Epoch 01240: val_loss did not improve\n",
      "0s - loss: 0.0090 - acc: 0.9968 - val_loss: 0.0623 - val_acc: 0.9835\n",
      "Epoch 1242/1536\n",
      "Epoch 01241: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9967 - val_loss: 0.0877 - val_acc: 0.9808\n",
      "Epoch 1243/1536\n",
      "Epoch 01242: val_loss did not improve\n",
      "0s - loss: 0.5974 - acc: 0.9285 - val_loss: 1.7694 - val_acc: 0.7978\n",
      "Epoch 1244/1536\n",
      "Epoch 01243: val_loss did not improve\n",
      "0s - loss: 0.6513 - acc: 0.9107 - val_loss: 0.0621 - val_acc: 0.9848\n",
      "Epoch 1245/1536\n",
      "Epoch 01244: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9958 - val_loss: 0.0602 - val_acc: 0.9858\n",
      "Epoch 1246/1536\n",
      "Epoch 01245: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9959 - val_loss: 0.0617 - val_acc: 0.9852\n",
      "Epoch 1247/1536\n",
      "Epoch 01246: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9968 - val_loss: 0.0610 - val_acc: 0.9857\n",
      "Epoch 1248/1536\n",
      "Epoch 01247: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9966 - val_loss: 0.0584 - val_acc: 0.9861\n",
      "Epoch 1249/1536\n",
      "Epoch 01248: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9972 - val_loss: 0.0604 - val_acc: 0.9853\n",
      "Epoch 1250/1536\n",
      "Epoch 01249: val_loss did not improve\n",
      "0s - loss: 0.0080 - acc: 0.9972 - val_loss: 0.0602 - val_acc: 0.9858\n",
      "Epoch 1251/1536\n",
      "Epoch 01250: val_loss did not improve\n",
      "0s - loss: 0.0097 - acc: 0.9968 - val_loss: 0.1142 - val_acc: 0.9755\n",
      "Epoch 1252/1536\n",
      "Epoch 01251: val_loss did not improve\n",
      "0s - loss: 1.5972 - acc: 0.8340 - val_loss: 0.1132 - val_acc: 0.9742\n",
      "Epoch 1253/1536\n",
      "Epoch 01252: val_loss did not improve\n",
      "0s - loss: 0.0274 - acc: 0.9920 - val_loss: 0.0692 - val_acc: 0.9834\n",
      "Epoch 1254/1536\n",
      "Epoch 01253: val_loss did not improve\n",
      "0s - loss: 0.0153 - acc: 0.9950 - val_loss: 0.0678 - val_acc: 0.9847\n",
      "Epoch 1255/1536\n",
      "Epoch 01254: val_loss did not improve\n",
      "0s - loss: 0.0121 - acc: 0.9961 - val_loss: 0.0613 - val_acc: 0.9851\n",
      "Epoch 1256/1536\n",
      "Epoch 01255: val_loss did not improve\n",
      "0s - loss: 0.0094 - acc: 0.9967 - val_loss: 0.0613 - val_acc: 0.9854\n",
      "Epoch 1257/1536\n",
      "Epoch 01256: val_loss did not improve\n",
      "0s - loss: 0.0097 - acc: 0.9968 - val_loss: 0.0679 - val_acc: 0.9837\n",
      "Epoch 1258/1536\n",
      "Epoch 01257: val_loss did not improve\n",
      "0s - loss: 0.0110 - acc: 0.9960 - val_loss: 0.0889 - val_acc: 0.9796\n",
      "Epoch 1259/1536\n",
      "Epoch 01258: val_loss did not improve\n",
      "0s - loss: 0.6789 - acc: 0.9188 - val_loss: 1.5503 - val_acc: 0.8158\n",
      "Epoch 1260/1536\n",
      "Epoch 01259: val_loss did not improve\n",
      "0s - loss: 0.5750 - acc: 0.9243 - val_loss: 0.0702 - val_acc: 0.9844\n",
      "Epoch 1261/1536\n",
      "Epoch 01260: val_loss did not improve\n",
      "0s - loss: 0.0131 - acc: 0.9954 - val_loss: 0.0686 - val_acc: 0.9853\n",
      "Epoch 1262/1536\n",
      "Epoch 01261: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9964 - val_loss: 0.0632 - val_acc: 0.9858\n",
      "Epoch 1263/1536\n",
      "Epoch 01262: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9962 - val_loss: 0.0646 - val_acc: 0.9868\n",
      "Epoch 1264/1536\n",
      "Epoch 01263: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9971 - val_loss: 0.0664 - val_acc: 0.9841\n",
      "Epoch 1265/1536\n",
      "Epoch 01264: val_loss did not improve\n",
      "0s - loss: 0.0097 - acc: 0.9966 - val_loss: 0.1062 - val_acc: 0.9774\n",
      "Epoch 1266/1536\n",
      "Epoch 01265: val_loss did not improve\n",
      "0s - loss: 0.4096 - acc: 0.9504 - val_loss: 0.4249 - val_acc: 0.9262\n",
      "Epoch 1267/1536\n",
      "Epoch 01266: val_loss did not improve\n",
      "0s - loss: 0.2865 - acc: 0.9572 - val_loss: 0.0687 - val_acc: 0.9845\n",
      "Epoch 1268/1536\n",
      "Epoch 01267: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9965 - val_loss: 0.0699 - val_acc: 0.9837\n",
      "Epoch 1269/1536\n",
      "Epoch 01268: val_loss did not improve\n",
      "0s - loss: 0.0093 - acc: 0.9968 - val_loss: 0.0631 - val_acc: 0.9852\n",
      "Epoch 1270/1536\n",
      "Epoch 01269: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9974 - val_loss: 0.0638 - val_acc: 0.9848\n",
      "Epoch 1271/1536\n",
      "Epoch 01270: val_loss did not improve\n",
      "0s - loss: 0.0060 - acc: 0.9979 - val_loss: 0.0631 - val_acc: 0.9841\n",
      "Epoch 1272/1536\n",
      "Epoch 01271: val_loss did not improve\n",
      "0s - loss: 0.0385 - acc: 0.9918 - val_loss: 0.7212 - val_acc: 0.9095\n",
      "Epoch 1273/1536\n",
      "Epoch 01272: val_loss did not improve\n",
      "0s - loss: 0.7082 - acc: 0.9039 - val_loss: 0.0895 - val_acc: 0.9816\n",
      "Epoch 1274/1536\n",
      "Epoch 01273: val_loss did not improve\n",
      "0s - loss: 0.0137 - acc: 0.9959 - val_loss: 0.0650 - val_acc: 0.9857\n",
      "Epoch 1275/1536\n",
      "Epoch 01274: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9968 - val_loss: 0.0607 - val_acc: 0.9863\n",
      "Epoch 1276/1536\n",
      "Epoch 01275: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9972 - val_loss: 0.0701 - val_acc: 0.9832\n",
      "Epoch 1277/1536\n",
      "Epoch 01276: val_loss did not improve\n",
      "0s - loss: 0.0087 - acc: 0.9973 - val_loss: 0.0769 - val_acc: 0.9836\n",
      "Epoch 1278/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01277: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9967 - val_loss: 0.0636 - val_acc: 0.9847\n",
      "Epoch 1279/1536\n",
      "Epoch 01278: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9943 - val_loss: 1.0425 - val_acc: 0.8950\n",
      "Epoch 1280/1536\n",
      "Epoch 01279: val_loss did not improve\n",
      "0s - loss: 2.4096 - acc: 0.7736 - val_loss: 0.2448 - val_acc: 0.9592\n",
      "Epoch 1281/1536\n",
      "Epoch 01280: val_loss did not improve\n",
      "0s - loss: 0.0663 - acc: 0.9857 - val_loss: 0.0781 - val_acc: 0.9817\n",
      "Epoch 1282/1536\n",
      "Epoch 01281: val_loss did not improve\n",
      "0s - loss: 0.0230 - acc: 0.9935 - val_loss: 0.0717 - val_acc: 0.9836\n",
      "Epoch 1283/1536\n",
      "Epoch 01282: val_loss did not improve\n",
      "0s - loss: 0.0170 - acc: 0.9945 - val_loss: 0.0695 - val_acc: 0.9841\n",
      "Epoch 1284/1536\n",
      "Epoch 01283: val_loss did not improve\n",
      "0s - loss: 0.0133 - acc: 0.9962 - val_loss: 0.0713 - val_acc: 0.9841\n",
      "Epoch 1285/1536\n",
      "Epoch 01284: val_loss did not improve\n",
      "0s - loss: 0.0119 - acc: 0.9959 - val_loss: 0.0780 - val_acc: 0.9822\n",
      "Epoch 1286/1536\n",
      "Epoch 01285: val_loss did not improve\n",
      "0s - loss: 0.0144 - acc: 0.9958 - val_loss: 0.0697 - val_acc: 0.9843\n",
      "Epoch 1287/1536\n",
      "Epoch 01286: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9966 - val_loss: 0.1329 - val_acc: 0.9742\n",
      "Epoch 1288/1536\n",
      "Epoch 01287: val_loss did not improve\n",
      "0s - loss: 0.6811 - acc: 0.9188 - val_loss: 0.5614 - val_acc: 0.9264\n",
      "Epoch 1289/1536\n",
      "Epoch 01288: val_loss did not improve\n",
      "0s - loss: 0.1169 - acc: 0.9801 - val_loss: 0.0685 - val_acc: 0.9837\n",
      "Epoch 1290/1536\n",
      "Epoch 01289: val_loss did not improve\n",
      "0s - loss: 0.0113 - acc: 0.9963 - val_loss: 0.0664 - val_acc: 0.9848\n",
      "Epoch 1291/1536\n",
      "Epoch 01290: val_loss did not improve\n",
      "0s - loss: 0.0109 - acc: 0.9965 - val_loss: 0.0652 - val_acc: 0.9838\n",
      "Epoch 1292/1536\n",
      "Epoch 01291: val_loss did not improve\n",
      "0s - loss: 0.0075 - acc: 0.9975 - val_loss: 0.0639 - val_acc: 0.9855\n",
      "Epoch 1293/1536\n",
      "Epoch 01292: val_loss did not improve\n",
      "0s - loss: 0.0084 - acc: 0.9974 - val_loss: 0.0835 - val_acc: 0.9828\n",
      "Epoch 1294/1536\n",
      "Epoch 01293: val_loss did not improve\n",
      "0s - loss: 0.6873 - acc: 0.9277 - val_loss: 0.8370 - val_acc: 0.8779\n",
      "Epoch 1295/1536\n",
      "Epoch 01294: val_loss did not improve\n",
      "0s - loss: 0.3318 - acc: 0.9471 - val_loss: 0.0736 - val_acc: 0.9849\n",
      "Epoch 1296/1536\n",
      "Epoch 01295: val_loss did not improve\n",
      "0s - loss: 0.0137 - acc: 0.9956 - val_loss: 0.0692 - val_acc: 0.9853\n",
      "Epoch 1297/1536\n",
      "Epoch 01296: val_loss did not improve\n",
      "0s - loss: 0.0109 - acc: 0.9966 - val_loss: 0.0771 - val_acc: 0.9834\n",
      "Epoch 1298/1536\n",
      "Epoch 01297: val_loss did not improve\n",
      "0s - loss: 0.0111 - acc: 0.9961 - val_loss: 0.0713 - val_acc: 0.9841\n",
      "Epoch 1299/1536\n",
      "Epoch 01298: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9965 - val_loss: 0.0690 - val_acc: 0.9851\n",
      "Epoch 1300/1536\n",
      "Epoch 01299: val_loss did not improve\n",
      "0s - loss: 0.0087 - acc: 0.9968 - val_loss: 0.0680 - val_acc: 0.9850\n",
      "Epoch 1301/1536\n",
      "Epoch 01300: val_loss did not improve\n",
      "0s - loss: 0.4160 - acc: 0.9508 - val_loss: 1.7752 - val_acc: 0.8058\n",
      "Epoch 1302/1536\n",
      "Epoch 01301: val_loss did not improve\n",
      "0s - loss: 1.0064 - acc: 0.8897 - val_loss: 0.0872 - val_acc: 0.9803\n",
      "Epoch 1303/1536\n",
      "Epoch 01302: val_loss did not improve\n",
      "0s - loss: 0.0240 - acc: 0.9928 - val_loss: 0.0758 - val_acc: 0.9839\n",
      "Epoch 1304/1536\n",
      "Epoch 01303: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9948 - val_loss: 0.0751 - val_acc: 0.9847\n",
      "Epoch 1305/1536\n",
      "Epoch 01304: val_loss did not improve\n",
      "0s - loss: 0.0127 - acc: 0.9957 - val_loss: 0.0710 - val_acc: 0.9853\n",
      "Epoch 1306/1536\n",
      "Epoch 01305: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9964 - val_loss: 0.0680 - val_acc: 0.9860\n",
      "Epoch 1307/1536\n",
      "Epoch 01306: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9966 - val_loss: 0.0759 - val_acc: 0.9835\n",
      "Epoch 1308/1536\n",
      "Epoch 01307: val_loss did not improve\n",
      "0s - loss: 0.0249 - acc: 0.9929 - val_loss: 0.5985 - val_acc: 0.9290\n",
      "Epoch 1309/1536\n",
      "Epoch 01308: val_loss did not improve\n",
      "0s - loss: 0.9012 - acc: 0.8901 - val_loss: 0.3612 - val_acc: 0.9373\n",
      "Epoch 1310/1536\n",
      "Epoch 01309: val_loss did not improve\n",
      "0s - loss: 0.1125 - acc: 0.9776 - val_loss: 0.0693 - val_acc: 0.9849\n",
      "Epoch 1311/1536\n",
      "Epoch 01310: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9955 - val_loss: 0.0689 - val_acc: 0.9852\n",
      "Epoch 1312/1536\n",
      "Epoch 01311: val_loss did not improve\n",
      "0s - loss: 0.0102 - acc: 0.9964 - val_loss: 0.0657 - val_acc: 0.9854\n",
      "Epoch 1313/1536\n",
      "Epoch 01312: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9972 - val_loss: 0.0717 - val_acc: 0.9846\n",
      "Epoch 1314/1536\n",
      "Epoch 01313: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9970 - val_loss: 0.0719 - val_acc: 0.9841\n",
      "Epoch 1315/1536\n",
      "Epoch 01314: val_loss did not improve\n",
      "0s - loss: 0.1060 - acc: 0.9847 - val_loss: 2.3268 - val_acc: 0.7753\n",
      "Epoch 1316/1536\n",
      "Epoch 01315: val_loss did not improve\n",
      "0s - loss: 1.1762 - acc: 0.8742 - val_loss: 0.0769 - val_acc: 0.9836\n",
      "Epoch 1317/1536\n",
      "Epoch 01316: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9952 - val_loss: 0.0706 - val_acc: 0.9841\n",
      "Epoch 1318/1536\n",
      "Epoch 01317: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9959 - val_loss: 0.0725 - val_acc: 0.9841\n",
      "Epoch 1319/1536\n",
      "Epoch 01318: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9964 - val_loss: 0.0739 - val_acc: 0.9825\n",
      "Epoch 1320/1536\n",
      "Epoch 01319: val_loss did not improve\n",
      "0s - loss: 0.0098 - acc: 0.9970 - val_loss: 0.0778 - val_acc: 0.9843\n",
      "Epoch 1321/1536\n",
      "Epoch 01320: val_loss did not improve\n",
      "0s - loss: 0.0294 - acc: 0.9922 - val_loss: 0.7924 - val_acc: 0.9229\n",
      "Epoch 1322/1536\n",
      "Epoch 01321: val_loss did not improve\n",
      "0s - loss: 0.6287 - acc: 0.9186 - val_loss: 0.1263 - val_acc: 0.9730\n",
      "Epoch 1323/1536\n",
      "Epoch 01322: val_loss did not improve\n",
      "0s - loss: 0.0194 - acc: 0.9941 - val_loss: 0.0669 - val_acc: 0.9860\n",
      "Epoch 1324/1536\n",
      "Epoch 01323: val_loss did not improve\n",
      "0s - loss: 0.0084 - acc: 0.9973 - val_loss: 0.0637 - val_acc: 0.9860\n",
      "Epoch 1325/1536\n",
      "Epoch 01324: val_loss did not improve\n",
      "0s - loss: 0.0073 - acc: 0.9973 - val_loss: 0.0759 - val_acc: 0.9829\n",
      "Epoch 1326/1536\n",
      "Epoch 01325: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9971 - val_loss: 0.0720 - val_acc: 0.9842\n",
      "Epoch 1327/1536\n",
      "Epoch 01326: val_loss did not improve\n",
      "0s - loss: 0.0691 - acc: 0.9854 - val_loss: 1.1495 - val_acc: 0.8925\n",
      "Epoch 1328/1536\n",
      "Epoch 01327: val_loss did not improve\n",
      "0s - loss: 1.1135 - acc: 0.8779 - val_loss: 0.2527 - val_acc: 0.9613\n",
      "Epoch 1329/1536\n",
      "Epoch 01328: val_loss did not improve\n",
      "0s - loss: 0.0585 - acc: 0.9882 - val_loss: 0.0757 - val_acc: 0.9846\n",
      "Epoch 1330/1536\n",
      "Epoch 01329: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9962 - val_loss: 0.0663 - val_acc: 0.9858\n",
      "Epoch 1331/1536\n",
      "Epoch 01330: val_loss did not improve\n",
      "0s - loss: 0.0110 - acc: 0.9965 - val_loss: 0.0641 - val_acc: 0.9864\n",
      "Epoch 1332/1536\n",
      "Epoch 01331: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9968 - val_loss: 0.0762 - val_acc: 0.9837\n",
      "Epoch 1333/1536\n",
      "Epoch 01332: val_loss did not improve\n",
      "0s - loss: 0.0110 - acc: 0.9969 - val_loss: 0.0734 - val_acc: 0.9831\n",
      "Epoch 1334/1536\n",
      "Epoch 01333: val_loss did not improve\n",
      "0s - loss: 0.0754 - acc: 0.9883 - val_loss: 0.9518 - val_acc: 0.9014\n",
      "Epoch 1335/1536\n",
      "Epoch 01334: val_loss did not improve\n",
      "0s - loss: 0.7775 - acc: 0.9036 - val_loss: 0.0741 - val_acc: 0.9858\n",
      "Epoch 1336/1536\n",
      "Epoch 01335: val_loss did not improve\n",
      "0s - loss: 0.0131 - acc: 0.9955 - val_loss: 0.0672 - val_acc: 0.9849\n",
      "Epoch 1337/1536\n",
      "Epoch 01336: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9965 - val_loss: 0.0676 - val_acc: 0.9844\n",
      "Epoch 1338/1536\n",
      "Epoch 01337: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9965 - val_loss: 0.0617 - val_acc: 0.9855\n",
      "Epoch 1339/1536\n",
      "Epoch 01338: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9973 - val_loss: 0.0729 - val_acc: 0.9834\n",
      "Epoch 1340/1536\n",
      "Epoch 01339: val_loss did not improve\n",
      "0s - loss: 0.2415 - acc: 0.9715 - val_loss: 0.9839 - val_acc: 0.8929\n",
      "Epoch 1341/1536\n",
      "Epoch 01340: val_loss did not improve\n",
      "0s - loss: 0.4719 - acc: 0.9331 - val_loss: 0.0750 - val_acc: 0.9831\n",
      "Epoch 1342/1536\n",
      "Epoch 01341: val_loss did not improve\n",
      "0s - loss: 0.0113 - acc: 0.9963 - val_loss: 0.0595 - val_acc: 0.9849\n",
      "Epoch 1343/1536\n",
      "Epoch 01342: val_loss did not improve\n",
      "0s - loss: 0.0072 - acc: 0.9975 - val_loss: 0.0595 - val_acc: 0.9849\n",
      "Epoch 1344/1536\n",
      "Epoch 01343: val_loss did not improve\n",
      "0s - loss: 0.0084 - acc: 0.9972 - val_loss: 0.0804 - val_acc: 0.9808\n",
      "Epoch 1345/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01344: val_loss did not improve\n",
      "0s - loss: 0.1483 - acc: 0.9762 - val_loss: 0.2277 - val_acc: 0.9608\n",
      "Epoch 1346/1536\n",
      "Epoch 01345: val_loss did not improve\n",
      "0s - loss: 0.2735 - acc: 0.9565 - val_loss: 0.1768 - val_acc: 0.9674\n",
      "Epoch 1347/1536\n",
      "Epoch 01346: val_loss did not improve\n",
      "0s - loss: 0.0316 - acc: 0.9927 - val_loss: 0.0584 - val_acc: 0.9855\n",
      "Epoch 1348/1536\n",
      "Epoch 01347: val_loss did not improve\n",
      "0s - loss: 0.0059 - acc: 0.9981 - val_loss: 0.0543 - val_acc: 0.9859\n",
      "Epoch 1349/1536\n",
      "Epoch 01348: val_loss did not improve\n",
      "0s - loss: 0.0057 - acc: 0.9983 - val_loss: 0.0629 - val_acc: 0.9857\n",
      "Epoch 1350/1536\n",
      "Epoch 01349: val_loss did not improve\n",
      "0s - loss: 0.0060 - acc: 0.9979 - val_loss: 0.0538 - val_acc: 0.9858\n",
      "Epoch 1351/1536\n",
      "Epoch 01350: val_loss did not improve\n",
      "0s - loss: 0.2683 - acc: 0.9733 - val_loss: 1.3912 - val_acc: 0.8706\n",
      "Epoch 1352/1536\n",
      "Epoch 01351: val_loss did not improve\n",
      "0s - loss: 0.8950 - acc: 0.8876 - val_loss: 0.0763 - val_acc: 0.9825\n",
      "Epoch 1353/1536\n",
      "Epoch 01352: val_loss did not improve\n",
      "0s - loss: 0.0132 - acc: 0.9954 - val_loss: 0.0625 - val_acc: 0.9849\n",
      "Epoch 1354/1536\n",
      "Epoch 01353: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9968 - val_loss: 0.0629 - val_acc: 0.9856\n",
      "Epoch 1355/1536\n",
      "Epoch 01354: val_loss did not improve\n",
      "0s - loss: 0.0094 - acc: 0.9967 - val_loss: 0.0635 - val_acc: 0.9857\n",
      "Epoch 1356/1536\n",
      "Epoch 01355: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9973 - val_loss: 0.0621 - val_acc: 0.9855\n",
      "Epoch 1357/1536\n",
      "Epoch 01356: val_loss did not improve\n",
      "0s - loss: 0.0072 - acc: 0.9977 - val_loss: 0.0657 - val_acc: 0.9846\n",
      "Epoch 1358/1536\n",
      "Epoch 01357: val_loss did not improve\n",
      "0s - loss: 0.0068 - acc: 0.9978 - val_loss: 0.0678 - val_acc: 0.9834\n",
      "Epoch 1359/1536\n",
      "Epoch 01358: val_loss did not improve\n",
      "0s - loss: 0.0492 - acc: 0.9899 - val_loss: 0.3922 - val_acc: 0.9304\n",
      "Epoch 1360/1536\n",
      "Epoch 01359: val_loss did not improve\n",
      "0s - loss: 0.5339 - acc: 0.9174 - val_loss: 0.0998 - val_acc: 0.9772\n",
      "Epoch 1361/1536\n",
      "Epoch 01360: val_loss did not improve\n",
      "0s - loss: 0.0149 - acc: 0.9954 - val_loss: 0.0672 - val_acc: 0.9843\n",
      "Epoch 1362/1536\n",
      "Epoch 01361: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9969 - val_loss: 0.0630 - val_acc: 0.9857\n",
      "Epoch 1363/1536\n",
      "Epoch 01362: val_loss did not improve\n",
      "0s - loss: 0.0077 - acc: 0.9975 - val_loss: 0.0698 - val_acc: 0.9829\n",
      "Epoch 1364/1536\n",
      "Epoch 01363: val_loss did not improve\n",
      "0s - loss: 0.0123 - acc: 0.9960 - val_loss: 0.0630 - val_acc: 0.9852\n",
      "Epoch 1365/1536\n",
      "Epoch 01364: val_loss did not improve\n",
      "0s - loss: 0.0320 - acc: 0.9945 - val_loss: 0.4928 - val_acc: 0.9393\n",
      "Epoch 1366/1536\n",
      "Epoch 01365: val_loss did not improve\n",
      "0s - loss: 0.4814 - acc: 0.9260 - val_loss: 0.4237 - val_acc: 0.9327\n",
      "Epoch 1367/1536\n",
      "Epoch 01366: val_loss did not improve\n",
      "0s - loss: 0.1050 - acc: 0.9787 - val_loss: 0.0650 - val_acc: 0.9849\n",
      "Epoch 1368/1536\n",
      "Epoch 01367: val_loss did not improve\n",
      "0s - loss: 0.0070 - acc: 0.9975 - val_loss: 0.0628 - val_acc: 0.9861\n",
      "Epoch 1369/1536\n",
      "Epoch 01368: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9971 - val_loss: 0.0630 - val_acc: 0.9850\n",
      "Epoch 1370/1536\n",
      "Epoch 01369: val_loss did not improve\n",
      "0s - loss: 0.0068 - acc: 0.9975 - val_loss: 0.0626 - val_acc: 0.9848\n",
      "Epoch 1371/1536\n",
      "Epoch 01370: val_loss did not improve\n",
      "0s - loss: 0.0070 - acc: 0.9976 - val_loss: 0.0869 - val_acc: 0.9800\n",
      "Epoch 1372/1536\n",
      "Epoch 01371: val_loss did not improve\n",
      "0s - loss: 0.6257 - acc: 0.9217 - val_loss: 0.3774 - val_acc: 0.9355\n",
      "Epoch 1373/1536\n",
      "Epoch 01372: val_loss did not improve\n",
      "0s - loss: 0.2555 - acc: 0.9556 - val_loss: 0.0643 - val_acc: 0.9848\n",
      "Epoch 1374/1536\n",
      "Epoch 01373: val_loss did not improve\n",
      "0s - loss: 0.0118 - acc: 0.9960 - val_loss: 0.0636 - val_acc: 0.9853\n",
      "Epoch 1375/1536\n",
      "Epoch 01374: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9970 - val_loss: 0.0702 - val_acc: 0.9827\n",
      "Epoch 1376/1536\n",
      "Epoch 01375: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9970 - val_loss: 0.0629 - val_acc: 0.9856\n",
      "Epoch 1377/1536\n",
      "Epoch 01376: val_loss did not improve\n",
      "0s - loss: 0.0071 - acc: 0.9974 - val_loss: 0.0663 - val_acc: 0.9854\n",
      "Epoch 1378/1536\n",
      "Epoch 01377: val_loss did not improve\n",
      "0s - loss: 0.2152 - acc: 0.9751 - val_loss: 0.5589 - val_acc: 0.9281\n",
      "Epoch 1379/1536\n",
      "Epoch 01378: val_loss did not improve\n",
      "0s - loss: 0.3113 - acc: 0.9480 - val_loss: 0.0734 - val_acc: 0.9810\n",
      "Epoch 1380/1536\n",
      "Epoch 01379: val_loss did not improve\n",
      "0s - loss: 0.0096 - acc: 0.9968 - val_loss: 0.0572 - val_acc: 0.9864\n",
      "Epoch 1381/1536\n",
      "Epoch 01380: val_loss did not improve\n",
      "0s - loss: 0.0071 - acc: 0.9975 - val_loss: 0.0551 - val_acc: 0.9855\n",
      "Epoch 1382/1536\n",
      "Epoch 01381: val_loss did not improve\n",
      "0s - loss: 0.0059 - acc: 0.9978 - val_loss: 0.0572 - val_acc: 0.9864\n",
      "Epoch 1383/1536\n",
      "Epoch 01382: val_loss did not improve\n",
      "0s - loss: 0.0058 - acc: 0.9981 - val_loss: 0.0862 - val_acc: 0.9797\n",
      "Epoch 1384/1536\n",
      "Epoch 01383: val_loss did not improve\n",
      "0s - loss: 0.8508 - acc: 0.9049 - val_loss: 0.1720 - val_acc: 0.9626\n",
      "Epoch 1385/1536\n",
      "Epoch 01384: val_loss did not improve\n",
      "0s - loss: 0.0464 - acc: 0.9879 - val_loss: 0.0594 - val_acc: 0.9849\n",
      "Epoch 1386/1536\n",
      "Epoch 01385: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9959 - val_loss: 0.0569 - val_acc: 0.9852\n",
      "Epoch 1387/1536\n",
      "Epoch 01386: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9963 - val_loss: 0.0656 - val_acc: 0.9841\n",
      "Epoch 1388/1536\n",
      "Epoch 01387: val_loss did not improve\n",
      "0s - loss: 0.0116 - acc: 0.9960 - val_loss: 0.0540 - val_acc: 0.9870\n",
      "Epoch 1389/1536\n",
      "Epoch 01388: val_loss did not improve\n",
      "0s - loss: 0.0084 - acc: 0.9972 - val_loss: 0.0587 - val_acc: 0.9857\n",
      "Epoch 1390/1536\n",
      "Epoch 01389: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9964 - val_loss: 0.1460 - val_acc: 0.9687\n",
      "Epoch 1391/1536\n",
      "Epoch 01390: val_loss did not improve\n",
      "0s - loss: 0.6634 - acc: 0.9091 - val_loss: 0.5483 - val_acc: 0.9109\n",
      "Epoch 1392/1536\n",
      "Epoch 01391: val_loss did not improve\n",
      "0s - loss: 0.1110 - acc: 0.9775 - val_loss: 0.0685 - val_acc: 0.9834\n",
      "Epoch 1393/1536\n",
      "Epoch 01392: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9964 - val_loss: 0.0627 - val_acc: 0.9846\n",
      "Epoch 1394/1536\n",
      "Epoch 01393: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9965 - val_loss: 0.0616 - val_acc: 0.9848\n",
      "Epoch 1395/1536\n",
      "Epoch 01394: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9972 - val_loss: 0.0625 - val_acc: 0.9840\n",
      "Epoch 1396/1536\n",
      "Epoch 01395: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9975 - val_loss: 0.0592 - val_acc: 0.9852\n",
      "Epoch 1397/1536\n",
      "Epoch 01396: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9975 - val_loss: 0.0871 - val_acc: 0.9785\n",
      "Epoch 1398/1536\n",
      "Epoch 01397: val_loss did not improve\n",
      "0s - loss: 0.8452 - acc: 0.8969 - val_loss: 0.6634 - val_acc: 0.8963\n",
      "Epoch 1399/1536\n",
      "Epoch 01398: val_loss did not improve\n",
      "0s - loss: 0.1648 - acc: 0.9697 - val_loss: 0.0674 - val_acc: 0.9840\n",
      "Epoch 1400/1536\n",
      "Epoch 01399: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9964 - val_loss: 0.0701 - val_acc: 0.9828\n",
      "Epoch 1401/1536\n",
      "Epoch 01400: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9967 - val_loss: 0.0648 - val_acc: 0.9841\n",
      "Epoch 1402/1536\n",
      "Epoch 01401: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9969 - val_loss: 0.0669 - val_acc: 0.9838\n",
      "Epoch 1403/1536\n",
      "Epoch 01402: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9966 - val_loss: 0.0683 - val_acc: 0.9832\n",
      "Epoch 1404/1536\n",
      "Epoch 01403: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9972 - val_loss: 0.0616 - val_acc: 0.9842\n",
      "Epoch 1405/1536\n",
      "Epoch 01404: val_loss did not improve\n",
      "0s - loss: 0.0079 - acc: 0.9974 - val_loss: 0.0722 - val_acc: 0.9818\n",
      "Epoch 1406/1536\n",
      "Epoch 01405: val_loss did not improve\n",
      "0s - loss: 0.6503 - acc: 0.9290 - val_loss: 3.5332 - val_acc: 0.6703\n",
      "Epoch 1407/1536\n",
      "Epoch 01406: val_loss did not improve\n",
      "0s - loss: 1.3704 - acc: 0.8586 - val_loss: 0.0819 - val_acc: 0.9815\n",
      "Epoch 1408/1536\n",
      "Epoch 01407: val_loss did not improve\n",
      "0s - loss: 0.0188 - acc: 0.9939 - val_loss: 0.0710 - val_acc: 0.9841\n",
      "Epoch 1409/1536\n",
      "Epoch 01408: val_loss did not improve\n",
      "0s - loss: 0.0164 - acc: 0.9947 - val_loss: 0.0655 - val_acc: 0.9851\n",
      "Epoch 1410/1536\n",
      "Epoch 01409: val_loss did not improve\n",
      "0s - loss: 0.0122 - acc: 0.9961 - val_loss: 0.0681 - val_acc: 0.9841\n",
      "Epoch 1411/1536\n",
      "Epoch 01410: val_loss did not improve\n",
      "0s - loss: 0.0125 - acc: 0.9956 - val_loss: 0.0637 - val_acc: 0.9855\n",
      "Epoch 1412/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01411: val_loss did not improve\n",
      "0s - loss: 0.0106 - acc: 0.9969 - val_loss: 0.0661 - val_acc: 0.9845\n",
      "Epoch 1413/1536\n",
      "Epoch 01412: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9962 - val_loss: 0.0665 - val_acc: 0.9855\n",
      "Epoch 1414/1536\n",
      "Epoch 01413: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9973 - val_loss: 0.0760 - val_acc: 0.9840\n",
      "Epoch 1415/1536\n",
      "Epoch 01414: val_loss did not improve\n",
      "0s - loss: 0.3120 - acc: 0.9595 - val_loss: 0.8999 - val_acc: 0.8724\n",
      "Epoch 1416/1536\n",
      "Epoch 01415: val_loss did not improve\n",
      "0s - loss: 0.4055 - acc: 0.9386 - val_loss: 0.0772 - val_acc: 0.9830\n",
      "Epoch 1417/1536\n",
      "Epoch 01416: val_loss did not improve\n",
      "0s - loss: 0.0115 - acc: 0.9961 - val_loss: 0.0673 - val_acc: 0.9841\n",
      "Epoch 1418/1536\n",
      "Epoch 01417: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9968 - val_loss: 0.0698 - val_acc: 0.9835\n",
      "Epoch 1419/1536\n",
      "Epoch 01418: val_loss did not improve\n",
      "0s - loss: 0.0099 - acc: 0.9968 - val_loss: 0.0669 - val_acc: 0.9839\n",
      "Epoch 1420/1536\n",
      "Epoch 01419: val_loss did not improve\n",
      "0s - loss: 0.0077 - acc: 0.9973 - val_loss: 0.0716 - val_acc: 0.9831\n",
      "Epoch 1421/1536\n",
      "Epoch 01420: val_loss did not improve\n",
      "0s - loss: 0.7636 - acc: 0.9177 - val_loss: 1.7403 - val_acc: 0.8123\n",
      "Epoch 1422/1536\n",
      "Epoch 01421: val_loss did not improve\n",
      "0s - loss: 0.5274 - acc: 0.9331 - val_loss: 0.0674 - val_acc: 0.9842\n",
      "Epoch 1423/1536\n",
      "Epoch 01422: val_loss did not improve\n",
      "0s - loss: 0.0125 - acc: 0.9957 - val_loss: 0.0644 - val_acc: 0.9850\n",
      "Epoch 1424/1536\n",
      "Epoch 01423: val_loss did not improve\n",
      "0s - loss: 0.0092 - acc: 0.9968 - val_loss: 0.0645 - val_acc: 0.9849\n",
      "Epoch 1425/1536\n",
      "Epoch 01424: val_loss did not improve\n",
      "0s - loss: 0.0111 - acc: 0.9964 - val_loss: 0.0628 - val_acc: 0.9858\n",
      "Epoch 1426/1536\n",
      "Epoch 01425: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9974 - val_loss: 0.0779 - val_acc: 0.9841\n",
      "Epoch 1427/1536\n",
      "Epoch 01426: val_loss did not improve\n",
      "0s - loss: 0.0097 - acc: 0.9965 - val_loss: 0.0649 - val_acc: 0.9854\n",
      "Epoch 1428/1536\n",
      "Epoch 01427: val_loss did not improve\n",
      "0s - loss: 0.0083 - acc: 0.9973 - val_loss: 0.1094 - val_acc: 0.9762\n",
      "Epoch 1429/1536\n",
      "Epoch 01428: val_loss did not improve\n",
      "0s - loss: 0.6532 - acc: 0.9149 - val_loss: 0.2659 - val_acc: 0.9516\n",
      "Epoch 1430/1536\n",
      "Epoch 01429: val_loss did not improve\n",
      "0s - loss: 0.0588 - acc: 0.9860 - val_loss: 0.0725 - val_acc: 0.9840\n",
      "Epoch 1431/1536\n",
      "Epoch 01430: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9965 - val_loss: 0.0699 - val_acc: 0.9850\n",
      "Epoch 1432/1536\n",
      "Epoch 01431: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9968 - val_loss: 0.0719 - val_acc: 0.9846\n",
      "Epoch 1433/1536\n",
      "Epoch 01432: val_loss did not improve\n",
      "0s - loss: 0.0090 - acc: 0.9969 - val_loss: 0.0740 - val_acc: 0.9845\n",
      "Epoch 1434/1536\n",
      "Epoch 01433: val_loss did not improve\n",
      "0s - loss: 0.0092 - acc: 0.9968 - val_loss: 0.0671 - val_acc: 0.9836\n",
      "Epoch 1435/1536\n",
      "Epoch 01434: val_loss did not improve\n",
      "0s - loss: 0.0081 - acc: 0.9972 - val_loss: 0.0615 - val_acc: 0.9854\n",
      "Epoch 1436/1536\n",
      "Epoch 01435: val_loss did not improve\n",
      "0s - loss: 0.1079 - acc: 0.9826 - val_loss: 1.9143 - val_acc: 0.8090\n",
      "Epoch 1437/1536\n",
      "Epoch 01436: val_loss did not improve\n",
      "0s - loss: 0.8534 - acc: 0.8968 - val_loss: 0.0817 - val_acc: 0.9816\n",
      "Epoch 1438/1536\n",
      "Epoch 01437: val_loss did not improve\n",
      "0s - loss: 0.0180 - acc: 0.9937 - val_loss: 0.0774 - val_acc: 0.9828\n",
      "Epoch 1439/1536\n",
      "Epoch 01438: val_loss did not improve\n",
      "0s - loss: 0.0128 - acc: 0.9958 - val_loss: 0.0782 - val_acc: 0.9829\n",
      "Epoch 1440/1536\n",
      "Epoch 01439: val_loss did not improve\n",
      "0s - loss: 0.0114 - acc: 0.9963 - val_loss: 0.0698 - val_acc: 0.9845\n",
      "Epoch 1441/1536\n",
      "Epoch 01440: val_loss did not improve\n",
      "0s - loss: 0.0080 - acc: 0.9974 - val_loss: 0.0707 - val_acc: 0.9853\n",
      "Epoch 1442/1536\n",
      "Epoch 01441: val_loss did not improve\n",
      "0s - loss: 0.1142 - acc: 0.9806 - val_loss: 0.6471 - val_acc: 0.8999\n",
      "Epoch 1443/1536\n",
      "Epoch 01442: val_loss did not improve\n",
      "0s - loss: 0.3950 - acc: 0.9385 - val_loss: 0.0651 - val_acc: 0.9843\n",
      "Epoch 1444/1536\n",
      "Epoch 01443: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9966 - val_loss: 0.0660 - val_acc: 0.9850\n",
      "Epoch 1445/1536\n",
      "Epoch 01444: val_loss did not improve\n",
      "0s - loss: 0.0090 - acc: 0.9967 - val_loss: 0.0730 - val_acc: 0.9823\n",
      "Epoch 1446/1536\n",
      "Epoch 01445: val_loss did not improve\n",
      "0s - loss: 0.0109 - acc: 0.9965 - val_loss: 0.0610 - val_acc: 0.9856\n",
      "Epoch 1447/1536\n",
      "Epoch 01446: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9951 - val_loss: 0.0905 - val_acc: 0.9785\n",
      "Epoch 1448/1536\n",
      "Epoch 01447: val_loss did not improve\n",
      "0s - loss: 0.4085 - acc: 0.9407 - val_loss: 0.9344 - val_acc: 0.8832\n",
      "Epoch 1449/1536\n",
      "Epoch 01448: val_loss did not improve\n",
      "0s - loss: 0.6126 - acc: 0.9181 - val_loss: 0.0681 - val_acc: 0.9856\n",
      "Epoch 1450/1536\n",
      "Epoch 01449: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9967 - val_loss: 0.0637 - val_acc: 0.9857\n",
      "Epoch 1451/1536\n",
      "Epoch 01450: val_loss did not improve\n",
      "0s - loss: 0.0096 - acc: 0.9967 - val_loss: 0.0652 - val_acc: 0.9856\n",
      "Epoch 1452/1536\n",
      "Epoch 01451: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9971 - val_loss: 0.0615 - val_acc: 0.9864\n",
      "Epoch 1453/1536\n",
      "Epoch 01452: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9975 - val_loss: 0.0645 - val_acc: 0.9856\n",
      "Epoch 1454/1536\n",
      "Epoch 01453: val_loss did not improve\n",
      "0s - loss: 0.0079 - acc: 0.9974 - val_loss: 0.0599 - val_acc: 0.9863\n",
      "Epoch 1455/1536\n",
      "Epoch 01454: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9973 - val_loss: 0.0856 - val_acc: 0.9820\n",
      "Epoch 1456/1536\n",
      "Epoch 01455: val_loss did not improve\n",
      "0s - loss: 0.4660 - acc: 0.9379 - val_loss: 1.5270 - val_acc: 0.8306\n",
      "Epoch 1457/1536\n",
      "Epoch 01456: val_loss did not improve\n",
      "0s - loss: 0.6375 - acc: 0.9223 - val_loss: 0.0674 - val_acc: 0.9850\n",
      "Epoch 1458/1536\n",
      "Epoch 01457: val_loss did not improve\n",
      "0s - loss: 0.0146 - acc: 0.9953 - val_loss: 0.0637 - val_acc: 0.9853\n",
      "Epoch 1459/1536\n",
      "Epoch 01458: val_loss did not improve\n",
      "0s - loss: 0.0120 - acc: 0.9962 - val_loss: 0.0658 - val_acc: 0.9847\n",
      "Epoch 1460/1536\n",
      "Epoch 01459: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9969 - val_loss: 0.0612 - val_acc: 0.9865\n",
      "Epoch 1461/1536\n",
      "Epoch 01460: val_loss did not improve\n",
      "0s - loss: 0.0098 - acc: 0.9966 - val_loss: 0.0570 - val_acc: 0.9861\n",
      "Epoch 1462/1536\n",
      "Epoch 01461: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9971 - val_loss: 0.0589 - val_acc: 0.9855\n",
      "Epoch 1463/1536\n",
      "Epoch 01462: val_loss did not improve\n",
      "0s - loss: 0.0081 - acc: 0.9974 - val_loss: 0.0916 - val_acc: 0.9795\n",
      "Epoch 1464/1536\n",
      "Epoch 01463: val_loss did not improve\n",
      "0s - loss: 1.1476 - acc: 0.8769 - val_loss: 0.2993 - val_acc: 0.9455\n",
      "Epoch 1465/1536\n",
      "Epoch 01464: val_loss did not improve\n",
      "0s - loss: 0.0685 - acc: 0.9840 - val_loss: 0.0681 - val_acc: 0.9846\n",
      "Epoch 1466/1536\n",
      "Epoch 01465: val_loss did not improve\n",
      "0s - loss: 0.0107 - acc: 0.9962 - val_loss: 0.0690 - val_acc: 0.9836\n",
      "Epoch 1467/1536\n",
      "Epoch 01466: val_loss did not improve\n",
      "0s - loss: 0.0111 - acc: 0.9961 - val_loss: 0.0704 - val_acc: 0.9825\n",
      "Epoch 1468/1536\n",
      "Epoch 01467: val_loss did not improve\n",
      "0s - loss: 0.0105 - acc: 0.9963 - val_loss: 0.0643 - val_acc: 0.9843\n",
      "Epoch 1469/1536\n",
      "Epoch 01468: val_loss did not improve\n",
      "0s - loss: 0.0086 - acc: 0.9971 - val_loss: 0.0622 - val_acc: 0.9840\n",
      "Epoch 1470/1536\n",
      "Epoch 01469: val_loss did not improve\n",
      "0s - loss: 0.0090 - acc: 0.9967 - val_loss: 0.0894 - val_acc: 0.9787\n",
      "Epoch 1471/1536\n",
      "Epoch 01470: val_loss did not improve\n",
      "0s - loss: 0.5412 - acc: 0.9288 - val_loss: 0.6021 - val_acc: 0.9105\n",
      "Epoch 1472/1536\n",
      "Epoch 01471: val_loss did not improve\n",
      "0s - loss: 0.2683 - acc: 0.9600 - val_loss: 0.0694 - val_acc: 0.9832\n",
      "Epoch 1473/1536\n",
      "Epoch 01472: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9960 - val_loss: 0.0640 - val_acc: 0.9846\n",
      "Epoch 1474/1536\n",
      "Epoch 01473: val_loss did not improve\n",
      "0s - loss: 0.0094 - acc: 0.9970 - val_loss: 0.0633 - val_acc: 0.9838\n",
      "Epoch 1475/1536\n",
      "Epoch 01474: val_loss did not improve\n",
      "0s - loss: 0.0078 - acc: 0.9973 - val_loss: 0.0718 - val_acc: 0.9831\n",
      "Epoch 1476/1536\n",
      "Epoch 01475: val_loss did not improve\n",
      "0s - loss: 0.1219 - acc: 0.9839 - val_loss: 0.1783 - val_acc: 0.9639\n",
      "Epoch 1477/1536\n",
      "Epoch 01476: val_loss did not improve\n",
      "0s - loss: 0.2795 - acc: 0.9494 - val_loss: 0.0753 - val_acc: 0.9812\n",
      "Epoch 1478/1536\n",
      "Epoch 01477: val_loss did not improve\n",
      "0s - loss: 0.0098 - acc: 0.9969 - val_loss: 0.0591 - val_acc: 0.9848\n",
      "Epoch 1479/1536\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 01478: val_loss did not improve\n",
      "0s - loss: 0.0066 - acc: 0.9979 - val_loss: 0.0646 - val_acc: 0.9838\n",
      "Epoch 1480/1536\n",
      "Epoch 01479: val_loss did not improve\n",
      "0s - loss: 0.0087 - acc: 0.9970 - val_loss: 0.0615 - val_acc: 0.9858\n",
      "Epoch 1481/1536\n",
      "Epoch 01480: val_loss did not improve\n",
      "0s - loss: 0.0080 - acc: 0.9972 - val_loss: 0.0784 - val_acc: 0.9813\n",
      "Epoch 1482/1536\n",
      "Epoch 01481: val_loss did not improve\n",
      "0s - loss: 0.4621 - acc: 0.9413 - val_loss: 0.6527 - val_acc: 0.8977\n",
      "Epoch 1483/1536\n",
      "Epoch 01482: val_loss did not improve\n",
      "0s - loss: 0.4940 - acc: 0.9231 - val_loss: 0.0696 - val_acc: 0.9837\n",
      "Epoch 1484/1536\n",
      "Epoch 01483: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9960 - val_loss: 0.0600 - val_acc: 0.9855\n",
      "Epoch 1485/1536\n",
      "Epoch 01484: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9969 - val_loss: 0.0603 - val_acc: 0.9851\n",
      "Epoch 1486/1536\n",
      "Epoch 01485: val_loss did not improve\n",
      "0s - loss: 0.0079 - acc: 0.9973 - val_loss: 0.0625 - val_acc: 0.9855\n",
      "Epoch 1487/1536\n",
      "Epoch 01486: val_loss did not improve\n",
      "0s - loss: 0.0074 - acc: 0.9975 - val_loss: 0.0573 - val_acc: 0.9854\n",
      "Epoch 1488/1536\n",
      "Epoch 01487: val_loss did not improve\n",
      "0s - loss: 0.0071 - acc: 0.9977 - val_loss: 0.0628 - val_acc: 0.9843\n",
      "Epoch 1489/1536\n",
      "Epoch 01488: val_loss did not improve\n",
      "0s - loss: 0.3224 - acc: 0.9599 - val_loss: 1.1946 - val_acc: 0.8576\n",
      "Epoch 1490/1536\n",
      "Epoch 01489: val_loss did not improve\n",
      "0s - loss: 0.4554 - acc: 0.9322 - val_loss: 0.0663 - val_acc: 0.9845\n",
      "Epoch 1491/1536\n",
      "Epoch 01490: val_loss did not improve\n",
      "0s - loss: 0.0097 - acc: 0.9969 - val_loss: 0.0642 - val_acc: 0.9851\n",
      "Epoch 1492/1536\n",
      "Epoch 01491: val_loss did not improve\n",
      "0s - loss: 0.0089 - acc: 0.9972 - val_loss: 0.0687 - val_acc: 0.9846\n",
      "Epoch 1493/1536\n",
      "Epoch 01492: val_loss did not improve\n",
      "0s - loss: 0.0091 - acc: 0.9971 - val_loss: 0.0597 - val_acc: 0.9852\n",
      "Epoch 1494/1536\n",
      "Epoch 01493: val_loss did not improve\n",
      "0s - loss: 0.0063 - acc: 0.9981 - val_loss: 0.0585 - val_acc: 0.9852\n",
      "Epoch 1495/1536\n",
      "Epoch 01494: val_loss did not improve\n",
      "0s - loss: 0.0064 - acc: 0.9978 - val_loss: 0.0587 - val_acc: 0.9850\n",
      "Epoch 1496/1536\n",
      "Epoch 01495: val_loss did not improve\n",
      "0s - loss: 0.0070 - acc: 0.9975 - val_loss: 0.0788 - val_acc: 0.9819\n",
      "Epoch 1497/1536\n",
      "Epoch 01496: val_loss did not improve\n",
      "0s - loss: 0.6041 - acc: 0.9321 - val_loss: 0.5933 - val_acc: 0.9150\n",
      "Epoch 1498/1536\n",
      "Epoch 01497: val_loss did not improve\n",
      "0s - loss: 0.2334 - acc: 0.9586 - val_loss: 0.0608 - val_acc: 0.9847\n",
      "Epoch 1499/1536\n",
      "Epoch 01498: val_loss did not improve\n",
      "0s - loss: 0.0109 - acc: 0.9963 - val_loss: 0.0654 - val_acc: 0.9833\n",
      "Epoch 1500/1536\n",
      "Epoch 01499: val_loss did not improve\n",
      "0s - loss: 0.0099 - acc: 0.9967 - val_loss: 0.0555 - val_acc: 0.9856\n",
      "Epoch 1501/1536\n",
      "Epoch 01500: val_loss did not improve\n",
      "0s - loss: 0.0080 - acc: 0.9971 - val_loss: 0.0674 - val_acc: 0.9836\n",
      "Epoch 1502/1536\n",
      "Epoch 01501: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9967 - val_loss: 0.0556 - val_acc: 0.9860\n",
      "Epoch 1503/1536\n",
      "Epoch 01502: val_loss did not improve\n",
      "0s - loss: 0.0077 - acc: 0.9974 - val_loss: 0.0754 - val_acc: 0.9811\n",
      "Epoch 1504/1536\n",
      "Epoch 01503: val_loss did not improve\n",
      "0s - loss: 0.2966 - acc: 0.9520 - val_loss: 1.3840 - val_acc: 0.8382\n",
      "Epoch 1505/1536\n",
      "Epoch 01504: val_loss did not improve\n",
      "0s - loss: 0.8900 - acc: 0.8869 - val_loss: 0.0652 - val_acc: 0.9841\n",
      "Epoch 1506/1536\n",
      "Epoch 01505: val_loss did not improve\n",
      "0s - loss: 0.0138 - acc: 0.9952 - val_loss: 0.0596 - val_acc: 0.9858\n",
      "Epoch 1507/1536\n",
      "Epoch 01506: val_loss did not improve\n",
      "0s - loss: 0.0104 - acc: 0.9967 - val_loss: 0.0565 - val_acc: 0.9862\n",
      "Epoch 1508/1536\n",
      "Epoch 01507: val_loss did not improve\n",
      "0s - loss: 0.0100 - acc: 0.9968 - val_loss: 0.0559 - val_acc: 0.9862\n",
      "Epoch 1509/1536\n",
      "Epoch 01508: val_loss did not improve\n",
      "0s - loss: 0.0101 - acc: 0.9969 - val_loss: 0.0557 - val_acc: 0.9861\n",
      "Epoch 1510/1536\n",
      "Epoch 01509: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9970 - val_loss: 0.0529 - val_acc: 0.9871\n",
      "Epoch 1511/1536\n",
      "Epoch 01510: val_loss did not improve\n",
      "0s - loss: 0.0070 - acc: 0.9974 - val_loss: 0.0736 - val_acc: 0.9821\n",
      "Epoch 1512/1536\n",
      "Epoch 01511: val_loss did not improve\n",
      "0s - loss: 0.6874 - acc: 0.9188 - val_loss: 0.5794 - val_acc: 0.9105\n",
      "Epoch 1513/1536\n",
      "Epoch 01512: val_loss did not improve\n",
      "0s - loss: 0.1288 - acc: 0.9758 - val_loss: 0.0607 - val_acc: 0.9851\n",
      "Epoch 1514/1536\n",
      "Epoch 01513: val_loss did not improve\n",
      "0s - loss: 0.0124 - acc: 0.9961 - val_loss: 0.0582 - val_acc: 0.9861\n",
      "Epoch 1515/1536\n",
      "Epoch 01514: val_loss did not improve\n",
      "0s - loss: 0.0108 - acc: 0.9967 - val_loss: 0.0581 - val_acc: 0.9862\n",
      "Epoch 1516/1536\n",
      "Epoch 01515: val_loss did not improve\n",
      "0s - loss: 0.0084 - acc: 0.9970 - val_loss: 0.0575 - val_acc: 0.9858\n",
      "Epoch 1517/1536\n",
      "Epoch 01516: val_loss did not improve\n",
      "0s - loss: 0.0085 - acc: 0.9973 - val_loss: 0.0578 - val_acc: 0.9855\n",
      "Epoch 1518/1536\n",
      "Epoch 01517: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9968 - val_loss: 0.0702 - val_acc: 0.9839\n",
      "Epoch 1519/1536\n",
      "Epoch 01518: val_loss did not improve\n",
      "0s - loss: 0.4847 - acc: 0.9393 - val_loss: 0.7632 - val_acc: 0.8908\n",
      "Epoch 1520/1536\n",
      "Epoch 01519: val_loss did not improve\n",
      "0s - loss: 0.2378 - acc: 0.9605 - val_loss: 0.0647 - val_acc: 0.9845\n",
      "Epoch 1521/1536\n",
      "Epoch 01520: val_loss did not improve\n",
      "0s - loss: 0.0103 - acc: 0.9967 - val_loss: 0.0603 - val_acc: 0.9844\n",
      "Epoch 1522/1536\n",
      "Epoch 01521: val_loss did not improve\n",
      "0s - loss: 0.0082 - acc: 0.9968 - val_loss: 0.0648 - val_acc: 0.9838\n",
      "Epoch 1523/1536\n",
      "Epoch 01522: val_loss did not improve\n",
      "0s - loss: 0.0072 - acc: 0.9975 - val_loss: 0.0581 - val_acc: 0.9856\n",
      "Epoch 1524/1536\n",
      "Epoch 01523: val_loss did not improve\n",
      "0s - loss: 0.0060 - acc: 0.9980 - val_loss: 0.0843 - val_acc: 0.9816\n",
      "Epoch 1525/1536\n",
      "Epoch 01524: val_loss did not improve\n",
      "0s - loss: 0.2263 - acc: 0.9612 - val_loss: 0.7293 - val_acc: 0.8947\n",
      "Epoch 1526/1536\n",
      "Epoch 01525: val_loss did not improve\n",
      "0s - loss: 0.4774 - acc: 0.9293 - val_loss: 0.0654 - val_acc: 0.9838\n",
      "Epoch 1527/1536\n",
      "Epoch 01526: val_loss did not improve\n",
      "0s - loss: 0.0088 - acc: 0.9969 - val_loss: 0.0611 - val_acc: 0.9842\n",
      "Epoch 1528/1536\n",
      "Epoch 01527: val_loss did not improve\n",
      "0s - loss: 0.0073 - acc: 0.9978 - val_loss: 0.0607 - val_acc: 0.9845\n",
      "Epoch 1529/1536\n",
      "Epoch 01528: val_loss did not improve\n",
      "0s - loss: 0.0071 - acc: 0.9976 - val_loss: 0.0588 - val_acc: 0.9853\n",
      "Epoch 1530/1536\n",
      "Epoch 01529: val_loss did not improve\n",
      "0s - loss: 0.0072 - acc: 0.9976 - val_loss: 0.0570 - val_acc: 0.9858\n",
      "Epoch 1531/1536\n",
      "Epoch 01530: val_loss did not improve\n",
      "0s - loss: 0.0051 - acc: 0.9983 - val_loss: 0.0678 - val_acc: 0.9840\n",
      "Epoch 1532/1536\n",
      "Epoch 01531: val_loss did not improve\n",
      "0s - loss: 0.4678 - acc: 0.9418 - val_loss: 0.8597 - val_acc: 0.8720\n",
      "Epoch 1533/1536\n",
      "Epoch 01532: val_loss did not improve\n",
      "0s - loss: 0.4288 - acc: 0.9330 - val_loss: 0.0654 - val_acc: 0.9823\n",
      "Epoch 1534/1536\n",
      "Epoch 01533: val_loss did not improve\n",
      "0s - loss: 0.0115 - acc: 0.9961 - val_loss: 0.0599 - val_acc: 0.9847\n",
      "Epoch 1535/1536\n",
      "Epoch 01534: val_loss did not improve\n",
      "0s - loss: 0.0095 - acc: 0.9972 - val_loss: 0.0570 - val_acc: 0.9856\n",
      "Epoch 1536/1536\n",
      "Epoch 01535: val_loss did not improve\n",
      "0s - loss: 0.0075 - acc: 0.9974 - val_loss: 0.0597 - val_acc: 0.9842\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<keras.callbacks.History at 0x7f9356161b00>"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "epochs = 1536\n",
    "batchsize = 8192\n",
    "model_details = 'resnet50_{}_batch{}_epoch{}'.format(model_name,batchsize,epochs)\n",
    "model_weights_path = '{}/{}.hdf5'.format(models_dir,model_details)\n",
    "\n",
    "checkpointer = ModelCheckpoint(filepath=model_weights_path, \n",
    "                               verbose=1, save_best_only=True)\n",
    "\n",
    "model.fit(features, targets, batch_size=batchsize, epochs=epochs,\n",
    "          validation_data=(features_valid, valid_targets),\n",
    "          callbacks=[checkpointer],\n",
    "          verbose=2, shuffle=True)          "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluating the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Test accuracy: 69.7152%\n"
     ]
    }
   ],
   "source": [
    "# Load the weights that yielded the best validation accuracy\n",
    "model.load_weights(model_weights_path)\n",
    "\n",
    "# Evaluate test accuracy\n",
    "score = model.evaluate(features_test, test_targets, verbose=0)\n",
    "accuracy = 100*score[1]\n",
    "\n",
    "# Print test accuracy\n",
    "print('Test accuracy: %.4f%%' % accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Confusion Matrix of Test Data\n",
    "\n",
    "After performing tests, a confusion matrix can be used to evaluate the model and it's accuracy for each ASL handshape/letter. Below we use the weight's from the last model defined and used above to generate our confusion matrix."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "12768/12782 [============================>.] - ETA: 0s"
     ]
    }
   ],
   "source": [
    "# Load the weights of latest model run\n",
    "model.load_weights(model_weights_path)\n",
    "\n",
    "test_predictions = model.predict_classes(features_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "sess = tf.Session()\n",
    "# Change one-hot encoding to just an integer representing letter\n",
    "un_hot_encode_targets = np.argmax(test_targets, axis=1)\n",
    "\n",
    "confusion_arr = sess.run(tf.confusion_matrix(un_hot_encode_targets, test_predictions, 24))\n",
    "# Get percents to compare with research paper benchmark\n",
    "confusion_arr_percents = confusion_arr/sum(confusion_arr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "A: 705\n",
      "B: 586\n",
      "C: 519\n",
      "D: 461\n",
      "E: 764\n",
      "F: 450\n",
      "G: 766\n",
      "H: 185\n",
      "I: 450\n",
      "K: 494\n",
      "L: 469\n",
      "M: 453\n",
      "N: 1239\n",
      "O: 380\n",
      "P: 526\n",
      "Q: 285\n",
      "R: 681\n",
      "S: 513\n",
      "T: 225\n",
      "U: 671\n",
      "V: 725\n",
      "W: 519\n",
      "X: 224\n",
      "Y: 492\n"
     ]
    }
   ],
   "source": [
    "# Note no 'J' & 'Z' were used\n",
    "letters = \"ABCDEFGHIKLMNOPQRSTUVWXY\"\n",
    "\n",
    "# Print out counts to see distribution of the number of letters tested against\n",
    "for elem in ['{}: {}'.format(letter,count) for letter,count in zip(letters,sum(confusion_arr))]:\n",
    "    print(elem)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save confusion matrix/array as CSVs for reference\n",
    "np.savetxt('confusion_matrices/confusionMatrix_{}_counts.csv'.format(model_details),confusion_arr)\n",
    "np.savetxt('confusion_matrices/confusionMatrix_{}_percent.csv'.format(model_details),confusion_arr_percents)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f9290cb7160>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABREAAARiCAYAAAAzw7DEAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMHgAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f9290cac9b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df_cm = pd.DataFrame(confusion_arr, index = [i for i in letters],\n",
    "                  columns = [i for i in letters])\n",
    "plt.figure(figsize = (25,20))\n",
    "sn.heatmap(df_cm, annot=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f928f9fbbe0>"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAxUAAAJCCAYAAAC24erxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\ngN8DPQUAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f92ea5f1630>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Percentages for confusion matrix\n",
    "# Multiplied by 100 to make it easier to read values\n",
    "df_cm = pd.DataFrame(confusion_arr_percents*100, index = [i for i in letters],\n",
    "                  columns = [i for i in letters])\n",
    "plt.figure(figsize = (15,10))\n",
    "sn.heatmap(df_cm, annot=True)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}