keras-u-net-starter-v3.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python2.7/dist-packages/h5py/__init__.py:34: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.7579927\n",
      "3.7836285\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "os.environ['PYTHONHASHSEED'] = '0'\n",
    "\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "tf.reset_default_graph()\n",
    "\n",
    "random_seed = 42\n",
    "from numpy.random import seed\n",
    "seed(random_seed)\n",
    "from tensorflow import set_random_seed\n",
    "set_random_seed(random_seed)\n",
    "import random\n",
    "random.seed(random_seed)\n",
    "#print(random.getstate())\n",
    "\n",
    "from keras import backend as K\n",
    "\n",
    "with tf.device('/gpu:0'):\n",
    "    c = tf.random_uniform([], -10, 10)\n",
    "    d = tf.random_uniform([], -10, 10)\n",
    "\n",
    "with tf.Session(config=tf.ConfigProto(log_device_placement=True)) as sess:\n",
    "    print(sess.run(c))\n",
    "    print(sess.run(d))\n",
    "    \n",
    "    \n",
    "session_conf = tf.ConfigProto(intra_op_parallelism_threads=1, inter_op_parallelism_threads=1)   \n",
    "sess = tf.Session(graph=tf.get_default_graph(), config=session_conf)    \n",
    "\n",
    "K.set_session(sess)    \n",
    "#print(sess)\n",
    "#config = tf.ConfigProto(\n",
    "#        device_count = {'GPU': 0}\n",
    "#    )\n",
    "#sess = tf.Session(config=config)\n",
    "# https://machinelearningmastery.com/reproducible-results-neural-networks-keras/\n",
    "# https://stackoverflow.com/questions/45230448/how-to-get-reproducible-result-when-running-keras-with-tensorflow-backend\n",
    "# https://stackoverflow.com/questions/46836857/results-not-reproducible-with-keras-and-tensorflow-in-python"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2.1.3\n"
     ]
    }
   ],
   "source": [
    "import os\n",
    "import sys\n",
    "import warnings\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from tqdm import tqdm\n",
    "from itertools import chain\n",
    "from skimage.io import imread, imshow, imread_collection, concatenate_images\n",
    "from skimage.transform import resize\n",
    "from skimage.morphology import label\n",
    "\n",
    "from keras.models import Model, load_model\n",
    "from keras.layers import Input\n",
    "from keras.layers.core import Lambda\n",
    "from keras.layers.convolutional import Conv2D, Conv2DTranspose\n",
    "from keras.layers.pooling import MaxPooling2D\n",
    "from keras.layers.merge import concatenate\n",
    "from keras.callbacks import EarlyStopping, ModelCheckpoint\n",
    "from keras.optimizers import SGD, RMSprop, Adam\n",
    "\n",
    "import keras\n",
    "print(keras.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set some parameters\n",
    "IMG_WIDTH = 128\n",
    "IMG_HEIGHT = 128\n",
    "IMG_CHANNELS = 3\n",
    "TRAIN_PATH = 'data/stage1_train/'\n",
    "TEST_PATH = 'data/stage1_test/'\n",
    "DATA_PATH = 'data/'\n",
    "\n",
    "warnings.filterwarnings('ignore', category=UserWarning, module='skimage')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "670\n",
      "65\n"
     ]
    }
   ],
   "source": [
    "# Get train and test IDs\n",
    "train_ids = next(os.walk(TRAIN_PATH))[1]\n",
    "test_ids = next(os.walk(TEST_PATH))[1]\n",
    "print(len(train_ids))\n",
    "print(len(test_ids))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "/home/chen/workspace/git_examples/kaggle-nuclei-challenge\n"
     ]
    }
   ],
   "source": [
    "current_dir = os.getcwd()\n",
    "print(current_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "import bcolz\n",
    "def save_array(fname, arr): c=bcolz.carray(arr, rootdir=fname, mode='w'); c.flush()\n",
    "def load_array(fname): return bcolz.open(fname)[:]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "def fit_model():\n",
    "    model = get_model_bn_do()\n",
    "    model.fit_generator(batches, batches.N, nb_epoch=1, verbose=0,\n",
    "                        validation_data=test_batches, nb_val_samples=test_batches.N)\n",
    "    model.optimizer.lr=0.1\n",
    "    model.fit_generator(batches, batches.N, nb_epoch=4, verbose=0,\n",
    "                        validation_data=test_batches, nb_val_samples=test_batches.N)\n",
    "    model.optimizer.lr=0.01\n",
    "    model.fit_generator(batches, batches.N, nb_epoch=12, verbose=0,\n",
    "                        validation_data=test_batches, nb_val_samples=test_batches.N)\n",
    "    model.optimizer.lr=0.001\n",
    "    model.fit_generator(batches, batches.N, nb_epoch=18, verbose=0,\n",
    "                        validation_data=test_batches, nb_val_samples=test_batches.N)\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Getting and resizing train images and masks ... \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 670/670 [00:50<00:00, 13.24it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Getting and resizing test images ... \n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "100%|██████████| 65/65 [00:00<00:00, 140.34it/s]"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Done!\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    }
   ],
   "source": [
    "# Get and resize train images and masks\n",
    "X_train = np.zeros((len(train_ids), IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS), dtype=np.uint8)\n",
    "Y_train = np.zeros((len(train_ids), IMG_HEIGHT, IMG_WIDTH, 1), dtype=np.bool)\n",
    "print('Getting and resizing train images and masks ... ')\n",
    "sys.stdout.flush()\n",
    "\n",
    "filenames = []\n",
    "for n, id_ in tqdm(enumerate(train_ids), total=len(train_ids)):\n",
    "    path = TRAIN_PATH + id_\n",
    "    img = imread(path + '/images/' + id_ + '.png')[:,:,:IMG_CHANNELS]\n",
    "    filenames.append(path + '/images/' + id_ + '.png')\n",
    "    img = resize(img, (IMG_HEIGHT, IMG_WIDTH), mode='constant', preserve_range=True)\n",
    "    X_train[n] = img\n",
    "    mask = np.zeros((IMG_HEIGHT, IMG_WIDTH, 1), dtype=np.bool)\n",
    "    for mask_file in next(os.walk(path + '/masks/'))[2]:\n",
    "        mask_ = imread(path + '/masks/' + mask_file)\n",
    "        mask_ = np.expand_dims(resize(mask_, (IMG_HEIGHT, IMG_WIDTH), mode='constant', \n",
    "                                      preserve_range=True), axis=-1)\n",
    "        mask = np.maximum(mask, mask_)\n",
    "    Y_train[n] = mask\n",
    "\n",
    "# Get and resize test images\n",
    "X_test = np.zeros((len(test_ids), IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS), dtype=np.uint8)\n",
    "sizes_test = []\n",
    "print('Getting and resizing test images ... ')\n",
    "sys.stdout.flush()\n",
    "for n, id_ in tqdm(enumerate(test_ids), total=len(test_ids)):\n",
    "    path = TEST_PATH + id_\n",
    "    img = imread(path + '/images/' + id_ + '.png')[:,:,:IMG_CHANNELS]\n",
    "    sizes_test.append([img.shape[0], img.shape[1]])\n",
    "    img = resize(img, (IMG_HEIGHT, IMG_WIDTH), mode='constant', preserve_range=True)\n",
    "    X_test[n] = img\n",
    "\n",
    "save_array(DATA_PATH + 'X_train.bc', X_train)\n",
    "save_array(DATA_PATH + 'Y_train.bc', Y_train)\n",
    "save_array(DATA_PATH + 'X_test.bc', X_test)\n",
    "save_array(DATA_PATH + 'filenames.bc', filenames)\n",
    "\n",
    "print('Done!')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = load_array(DATA_PATH + 'X_train.bc')\n",
    "Y_train = load_array(DATA_PATH + 'Y_train.bc')\n",
    "X_test = load_array(DATA_PATH + 'X_test.bc')\n",
    "filenames = load_array(DATA_PATH + 'filenames.bc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Counter({(256, 256): 334,\n",
       "         (320, 256): 112,\n",
       "         (347, 260): 5,\n",
       "         (360, 360): 91,\n",
       "         (640, 512): 13,\n",
       "         (696, 520): 92,\n",
       "         (1024, 1024): 16,\n",
       "         (1272, 603): 6,\n",
       "         (1388, 1040): 1})"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import PIL\n",
    "import collections\n",
    "sizes = [PIL.Image.open(f).size for f in filenames]\n",
    "collections.Counter(sizes)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python2.7/dist-packages/skimage/io/_plugins/matplotlib_plugin.py:51: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  out_of_range_float = (np.issubdtype(image.dtype, np.float) and\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "102\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb2693d210>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbbb1cec790>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Check if training data looks all right\n",
    "ix = np.random.randint(0, len(train_ids))\n",
    "print(ix)\n",
    "imshow(X_train[ix])\n",
    "plt.show()\n",
    "imshow(np.squeeze(Y_train[ix]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Define IoU metric\n",
    "def mean_iou(y_true, y_pred):\n",
    "    prec = []\n",
    "    for t in np.arange(0.5, 1.0, 0.05):\n",
    "        y_pred_ = tf.to_int32(y_pred > t)\n",
    "        score, up_opt = tf.metrics.mean_iou(y_true, y_pred_, 2)\n",
    "        K.get_session().run(tf.local_variables_initializer())\n",
    "        with tf.control_dependencies([up_opt]):\n",
    "            score = tf.identity(score)\n",
    "        prec.append(score)\n",
    "    return K.mean(K.stack(prec), axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Build U-Net model\n",
    "inputs = Input((IMG_HEIGHT, IMG_WIDTH, IMG_CHANNELS))\n",
    "s = Lambda(lambda x: x / 255) (inputs)\n",
    "\n",
    "c1 = Conv2D(8, (3, 3), activation='relu', padding='same') (s)\n",
    "c1 = Conv2D(8, (3, 3), activation='relu', padding='same') (c1)\n",
    "p1 = MaxPooling2D((2, 2)) (c1)\n",
    "\n",
    "c2 = Conv2D(16, (3, 3), activation='relu', padding='same') (p1)\n",
    "c2 = Conv2D(16, (3, 3), activation='relu', padding='same') (c2)\n",
    "p2 = MaxPooling2D((2, 2)) (c2)\n",
    "\n",
    "c3 = Conv2D(32, (3, 3), activation='relu', padding='same') (p2)\n",
    "c3 = Conv2D(32, (3, 3), activation='relu', padding='same') (c3)\n",
    "p3 = MaxPooling2D((2, 2)) (c3)\n",
    "\n",
    "c4 = Conv2D(64, (3, 3), activation='relu', padding='same') (p3)\n",
    "c4 = Conv2D(64, (3, 3), activation='relu', padding='same') (c4)\n",
    "p4 = MaxPooling2D(pool_size=(2, 2)) (c4)\n",
    "\n",
    "c5 = Conv2D(128, (3, 3), activation='relu', padding='same') (p4)\n",
    "c5 = Conv2D(128, (3, 3), activation='relu', padding='same') (c5)\n",
    "\n",
    "u6 = Conv2DTranspose(64, (2, 2), strides=(2, 2), padding='same') (c5)\n",
    "u6 = concatenate([u6, c4])\n",
    "c6 = Conv2D(64, (3, 3), activation='relu', padding='same') (u6)\n",
    "c6 = Conv2D(64, (3, 3), activation='relu', padding='same') (c6)\n",
    "\n",
    "u7 = Conv2DTranspose(32, (2, 2), strides=(2, 2), padding='same') (c6)\n",
    "u7 = concatenate([u7, c3])\n",
    "c7 = Conv2D(32, (3, 3), activation='relu', padding='same') (u7)\n",
    "c7 = Conv2D(32, (3, 3), activation='relu', padding='same') (c7)\n",
    "\n",
    "u8 = Conv2DTranspose(16, (2, 2), strides=(2, 2), padding='same') (c7)\n",
    "u8 = concatenate([u8, c2])\n",
    "c8 = Conv2D(16, (3, 3), activation='relu', padding='same') (u8)\n",
    "c8 = Conv2D(16, (3, 3), activation='relu', padding='same') (c8)\n",
    "\n",
    "u9 = Conv2DTranspose(8, (2, 2), strides=(2, 2), padding='same') (c8)\n",
    "u9 = concatenate([u9, c1], axis=3)\n",
    "c9 = Conv2D(8, (3, 3), activation='relu', padding='same') (u9)\n",
    "c9 = Conv2D(8, (3, 3), activation='relu', padding='same') (c9)\n",
    "\n",
    "outputs = Conv2D(1, (1, 1), activation='sigmoid') (c9)\n",
    "\n",
    "model = Model(inputs=[inputs], outputs=[outputs])\n",
    "model.compile(optimizer='adam', loss='binary_crossentropy', metrics=[mean_iou])\n",
    "#model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 603 samples, validate on 67 samples\n",
      "Epoch 1/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.5708 - mean_iou: 0.4157\n",
      "Epoch 00001: val_loss improved from inf to 0.44784, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 3s 5ms/step - loss: 0.5661 - mean_iou: 0.4160 - val_loss: 0.4478 - val_mean_iou: 0.4246\n",
      "Epoch 2/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.3559 - mean_iou: 0.4247\n",
      "Epoch 00002: val_loss improved from 0.44784 to 0.27703, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.3542 - mean_iou: 0.4247 - val_loss: 0.2770 - val_mean_iou: 0.4251\n",
      "Epoch 3/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.2245 - mean_iou: 0.4418\n",
      "Epoch 00003: val_loss improved from 0.27703 to 0.20388, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.2220 - mean_iou: 0.4426 - val_loss: 0.2039 - val_mean_iou: 0.4712\n",
      "Epoch 4/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.1825 - mean_iou: 0.4963\n",
      "Epoch 00004: val_loss improved from 0.20388 to 0.18411, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.1811 - mean_iou: 0.4969 - val_loss: 0.1841 - val_mean_iou: 0.5156\n",
      "Epoch 5/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.1444 - mean_iou: 0.5338\n",
      "Epoch 00005: val_loss improved from 0.18411 to 0.16946, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.1449 - mean_iou: 0.5343 - val_loss: 0.1695 - val_mean_iou: 0.5533\n",
      "Epoch 6/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.1231 - mean_iou: 0.5690\n",
      "Epoch 00006: val_loss improved from 0.16946 to 0.13804, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.1228 - mean_iou: 0.5694 - val_loss: 0.1380 - val_mean_iou: 0.5854\n",
      "Epoch 7/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.1081 - mean_iou: 0.5996\n",
      "Epoch 00007: val_loss improved from 0.13804 to 0.12698, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.1078 - mean_iou: 0.5999 - val_loss: 0.1270 - val_mean_iou: 0.6136\n",
      "Epoch 8/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0992 - mean_iou: 0.6261\n",
      "Epoch 00008: val_loss improved from 0.12698 to 0.11991, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0988 - mean_iou: 0.6264 - val_loss: 0.1199 - val_mean_iou: 0.6372\n",
      "Epoch 9/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0976 - mean_iou: 0.6463\n",
      "Epoch 00009: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0981 - mean_iou: 0.6465 - val_loss: 0.1238 - val_mean_iou: 0.6557\n",
      "Epoch 10/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0899 - mean_iou: 0.6637\n",
      "Epoch 00010: val_loss improved from 0.11991 to 0.11288, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0907 - mean_iou: 0.6640 - val_loss: 0.1129 - val_mean_iou: 0.6720\n",
      "Epoch 11/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0882 - mean_iou: 0.6789\n",
      "Epoch 00011: val_loss improved from 0.11288 to 0.10895, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0881 - mean_iou: 0.6791 - val_loss: 0.1090 - val_mean_iou: 0.6855\n",
      "Epoch 12/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0865 - mean_iou: 0.6914\n",
      "Epoch 00012: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0850 - mean_iou: 0.6916 - val_loss: 0.1137 - val_mean_iou: 0.6972\n",
      "Epoch 13/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0845 - mean_iou: 0.7024\n",
      "Epoch 00013: val_loss improved from 0.10895 to 0.10830, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0840 - mean_iou: 0.7025 - val_loss: 0.1083 - val_mean_iou: 0.7076\n",
      "Epoch 14/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0797 - mean_iou: 0.7123\n",
      "Epoch 00014: val_loss improved from 0.10830 to 0.10420, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0803 - mean_iou: 0.7124 - val_loss: 0.1042 - val_mean_iou: 0.7170\n",
      "Epoch 15/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0798 - mean_iou: 0.7211\n",
      "Epoch 00015: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0799 - mean_iou: 0.7212 - val_loss: 0.1044 - val_mean_iou: 0.7251\n",
      "Epoch 16/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0797 - mean_iou: 0.7286\n",
      "Epoch 00016: val_loss improved from 0.10420 to 0.09750, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0799 - mean_iou: 0.7287 - val_loss: 0.0975 - val_mean_iou: 0.7323\n",
      "Epoch 17/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0756 - mean_iou: 0.7355\n",
      "Epoch 00017: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0756 - mean_iou: 0.7356 - val_loss: 0.1005 - val_mean_iou: 0.7391\n",
      "Epoch 18/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0752 - mean_iou: 0.7421\n",
      "Epoch 00018: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0755 - mean_iou: 0.7422 - val_loss: 0.1042 - val_mean_iou: 0.7451\n",
      "Epoch 19/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0738 - mean_iou: 0.7476\n",
      "Epoch 00019: val_loss improved from 0.09750 to 0.09456, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0738 - mean_iou: 0.7477 - val_loss: 0.0946 - val_mean_iou: 0.7505\n",
      "Epoch 20/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0760 - mean_iou: 0.7528\n",
      "Epoch 00020: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0751 - mean_iou: 0.7529 - val_loss: 0.0960 - val_mean_iou: 0.7555\n",
      "Epoch 21/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0735 - mean_iou: 0.7578\n",
      "Epoch 00021: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0730 - mean_iou: 0.7579 - val_loss: 0.0947 - val_mean_iou: 0.7602\n",
      "Epoch 22/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0723 - mean_iou: 0.7625\n",
      "Epoch 00022: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0718 - mean_iou: 0.7625 - val_loss: 0.0951 - val_mean_iou: 0.7646\n",
      "Epoch 23/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0706 - mean_iou: 0.7665\n",
      "Epoch 00023: val_loss improved from 0.09456 to 0.09404, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0709 - mean_iou: 0.7666 - val_loss: 0.0940 - val_mean_iou: 0.7686\n",
      "Epoch 24/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0713 - mean_iou: 0.7704\n",
      "Epoch 00024: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0716 - mean_iou: 0.7705 - val_loss: 0.1003 - val_mean_iou: 0.7722\n",
      "Epoch 25/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0731 - mean_iou: 0.7738\n",
      "Epoch 00025: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0731 - mean_iou: 0.7738 - val_loss: 0.1021 - val_mean_iou: 0.7755\n",
      "Epoch 26/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0715 - mean_iou: 0.7772\n",
      "Epoch 00026: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0714 - mean_iou: 0.7772 - val_loss: 0.1001 - val_mean_iou: 0.7787\n",
      "Epoch 27/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0704 - mean_iou: 0.7800\n",
      "Epoch 00027: val_loss improved from 0.09404 to 0.08803, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0707 - mean_iou: 0.7800 - val_loss: 0.0880 - val_mean_iou: 0.7816\n",
      "Epoch 28/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0692 - mean_iou: 0.7830\n",
      "Epoch 00028: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0689 - mean_iou: 0.7830 - val_loss: 0.0932 - val_mean_iou: 0.7845\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 29/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0689 - mean_iou: 0.7858\n",
      "Epoch 00029: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0687 - mean_iou: 0.7859 - val_loss: 0.0916 - val_mean_iou: 0.7872\n",
      "Epoch 30/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0678 - mean_iou: 0.7885\n",
      "Epoch 00030: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0679 - mean_iou: 0.7886 - val_loss: 0.0901 - val_mean_iou: 0.7898\n",
      "Epoch 31/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0682 - mean_iou: 0.7909\n",
      "Epoch 00031: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0677 - mean_iou: 0.7910 - val_loss: 0.0922 - val_mean_iou: 0.7922\n",
      "Epoch 32/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0671 - mean_iou: 0.7933\n",
      "Epoch 00032: val_loss improved from 0.08803 to 0.08684, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0662 - mean_iou: 0.7933 - val_loss: 0.0868 - val_mean_iou: 0.7944\n",
      "Epoch 33/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0653 - mean_iou: 0.7955\n",
      "Epoch 00033: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0655 - mean_iou: 0.7956 - val_loss: 0.0896 - val_mean_iou: 0.7967\n",
      "Epoch 34/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0648 - mean_iou: 0.7977\n",
      "Epoch 00034: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0655 - mean_iou: 0.7978 - val_loss: 0.0919 - val_mean_iou: 0.7988\n",
      "Epoch 35/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0671 - mean_iou: 0.7997\n",
      "Epoch 00035: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0676 - mean_iou: 0.7998 - val_loss: 0.0882 - val_mean_iou: 0.8007\n",
      "Epoch 36/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0655 - mean_iou: 0.8016\n",
      "Epoch 00036: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0653 - mean_iou: 0.8016 - val_loss: 0.0874 - val_mean_iou: 0.8026\n",
      "Epoch 37/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0642 - mean_iou: 0.8034\n",
      "Epoch 00037: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0643 - mean_iou: 0.8034 - val_loss: 0.0927 - val_mean_iou: 0.8044\n",
      "Epoch 38/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0636 - mean_iou: 0.8052\n",
      "Epoch 00038: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0635 - mean_iou: 0.8053 - val_loss: 0.0947 - val_mean_iou: 0.8061\n",
      "Epoch 39/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0649 - mean_iou: 0.8068\n",
      "Epoch 00039: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0649 - mean_iou: 0.8068 - val_loss: 0.0908 - val_mean_iou: 0.8077\n",
      "Epoch 40/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0634 - mean_iou: 0.8085\n",
      "Epoch 00040: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0637 - mean_iou: 0.8085 - val_loss: 0.0875 - val_mean_iou: 0.8093\n",
      "Epoch 41/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0637 - mean_iou: 0.8100\n",
      "Epoch 00041: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0636 - mean_iou: 0.8101 - val_loss: 0.0894 - val_mean_iou: 0.8108\n",
      "Epoch 42/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0630 - mean_iou: 0.8116\n",
      "Epoch 00042: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0627 - mean_iou: 0.8116 - val_loss: 0.0870 - val_mean_iou: 0.8123\n",
      "Epoch 43/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0649 - mean_iou: 0.8130\n",
      "Epoch 00043: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0643 - mean_iou: 0.8130 - val_loss: 0.1051 - val_mean_iou: 0.8137\n",
      "Epoch 44/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0614 - mean_iou: 0.8144\n",
      "Epoch 00044: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0609 - mean_iou: 0.8144 - val_loss: 0.0936 - val_mean_iou: 0.8151\n",
      "Epoch 45/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0621 - mean_iou: 0.8157\n",
      "Epoch 00045: val_loss improved from 0.08684 to 0.08645, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0631 - mean_iou: 0.8157 - val_loss: 0.0864 - val_mean_iou: 0.8164\n",
      "Epoch 46/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0593 - mean_iou: 0.8170\n",
      "Epoch 00046: val_loss improved from 0.08645 to 0.08400, saving model to model-tmp-v3.h5\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0598 - mean_iou: 0.8170 - val_loss: 0.0840 - val_mean_iou: 0.8177\n",
      "Epoch 47/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0584 - mean_iou: 0.8184\n",
      "Epoch 00047: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0593 - mean_iou: 0.8184 - val_loss: 0.0976 - val_mean_iou: 0.8191\n",
      "Epoch 48/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0584 - mean_iou: 0.8197\n",
      "Epoch 00048: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0592 - mean_iou: 0.8197 - val_loss: 0.0895 - val_mean_iou: 0.8203\n",
      "Epoch 49/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0573 - mean_iou: 0.8210\n",
      "Epoch 00049: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0571 - mean_iou: 0.8210 - val_loss: 0.0910 - val_mean_iou: 0.8217\n",
      "Epoch 50/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0588 - mean_iou: 0.8223\n",
      "Epoch 00050: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0593 - mean_iou: 0.8223 - val_loss: 0.0947 - val_mean_iou: 0.8229\n",
      "Epoch 51/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0578 - mean_iou: 0.8233\n",
      "Epoch 00051: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0581 - mean_iou: 0.8234 - val_loss: 0.0857 - val_mean_iou: 0.8240\n",
      "Epoch 52/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0561 - mean_iou: 0.8246\n",
      "Epoch 00052: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0556 - mean_iou: 0.8246 - val_loss: 0.0935 - val_mean_iou: 0.8252\n",
      "Epoch 53/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0549 - mean_iou: 0.8257\n",
      "Epoch 00053: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0556 - mean_iou: 0.8258 - val_loss: 0.0892 - val_mean_iou: 0.8264\n",
      "Epoch 54/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0554 - mean_iou: 0.8269\n",
      "Epoch 00054: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0547 - mean_iou: 0.8269 - val_loss: 0.0965 - val_mean_iou: 0.8275\n",
      "Epoch 55/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0537 - mean_iou: 0.8280\n",
      "Epoch 00055: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0541 - mean_iou: 0.8280 - val_loss: 0.0863 - val_mean_iou: 0.8286\n",
      "Epoch 56/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0524 - mean_iou: 0.8292\n",
      "Epoch 00056: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0530 - mean_iou: 0.8292 - val_loss: 0.0852 - val_mean_iou: 0.8297\n",
      "Epoch 57/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0531 - mean_iou: 0.8303\n",
      "Epoch 00057: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0535 - mean_iou: 0.8303 - val_loss: 0.0868 - val_mean_iou: 0.8308\n",
      "Epoch 58/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0538 - mean_iou: 0.8313\n",
      "Epoch 00058: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0541 - mean_iou: 0.8313 - val_loss: 0.0918 - val_mean_iou: 0.8318\n",
      "Epoch 59/100\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "584/603 [============================>.] - ETA: 0s - loss: 0.0524 - mean_iou: 0.8323\n",
      "Epoch 00059: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0518 - mean_iou: 0.8323 - val_loss: 0.0904 - val_mean_iou: 0.8329\n",
      "Epoch 60/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0510 - mean_iou: 0.8334\n",
      "Epoch 00060: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0506 - mean_iou: 0.8334 - val_loss: 0.0911 - val_mean_iou: 0.8339\n",
      "Epoch 61/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0515 - mean_iou: 0.8343\n",
      "Epoch 00061: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0514 - mean_iou: 0.8344 - val_loss: 0.0968 - val_mean_iou: 0.8349\n",
      "Epoch 62/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0515 - mean_iou: 0.8353\n",
      "Epoch 00062: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0511 - mean_iou: 0.8353 - val_loss: 0.0977 - val_mean_iou: 0.8358\n",
      "Epoch 63/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0486 - mean_iou: 0.8363\n",
      "Epoch 00063: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0484 - mean_iou: 0.8363 - val_loss: 0.0968 - val_mean_iou: 0.8368\n",
      "Epoch 64/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0476 - mean_iou: 0.8373\n",
      "Epoch 00064: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0480 - mean_iou: 0.8373 - val_loss: 0.0909 - val_mean_iou: 0.8378\n",
      "Epoch 65/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0492 - mean_iou: 0.8382\n",
      "Epoch 00065: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0494 - mean_iou: 0.8382 - val_loss: 0.0957 - val_mean_iou: 0.8387\n",
      "Epoch 66/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0488 - mean_iou: 0.8391\n",
      "Epoch 00066: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0488 - mean_iou: 0.8391 - val_loss: 0.1007 - val_mean_iou: 0.8396\n",
      "Epoch 67/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0491 - mean_iou: 0.8400\n",
      "Epoch 00067: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0494 - mean_iou: 0.8400 - val_loss: 0.0969 - val_mean_iou: 0.8404\n",
      "Epoch 68/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0468 - mean_iou: 0.8408\n",
      "Epoch 00068: val_loss did not improve\n",
      "603/603 [==============================] - 2s 2ms/step - loss: 0.0469 - mean_iou: 0.8409 - val_loss: 0.0906 - val_mean_iou: 0.8413\n",
      "Epoch 69/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0455 - mean_iou: 0.8417\n",
      "Epoch 00069: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0454 - mean_iou: 0.8417 - val_loss: 0.0954 - val_mean_iou: 0.8422\n",
      "Epoch 70/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0436 - mean_iou: 0.8426\n",
      "Epoch 00070: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0438 - mean_iou: 0.8426 - val_loss: 0.0922 - val_mean_iou: 0.8431\n",
      "Epoch 71/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0442 - mean_iou: 0.8435\n",
      "Epoch 00071: val_loss did not improve\n",
      "603/603 [==============================] - 2s 3ms/step - loss: 0.0441 - mean_iou: 0.8435 - val_loss: 0.0915 - val_mean_iou: 0.8440\n",
      "Epoch 72/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0423 - mean_iou: 0.8444\n",
      "Epoch 00072: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0422 - mean_iou: 0.8444 - val_loss: 0.0968 - val_mean_iou: 0.8449\n",
      "Epoch 73/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0417 - mean_iou: 0.8453\n",
      "Epoch 00073: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0417 - mean_iou: 0.8453 - val_loss: 0.0992 - val_mean_iou: 0.8458\n",
      "Epoch 74/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0410 - mean_iou: 0.8462\n",
      "Epoch 00074: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0409 - mean_iou: 0.8462 - val_loss: 0.0971 - val_mean_iou: 0.8467\n",
      "Epoch 75/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0401 - mean_iou: 0.8471\n",
      "Epoch 00075: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0406 - mean_iou: 0.8471 - val_loss: 0.0947 - val_mean_iou: 0.8475\n",
      "Epoch 76/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0400 - mean_iou: 0.8479\n",
      "Epoch 00076: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0397 - mean_iou: 0.8480 - val_loss: 0.0974 - val_mean_iou: 0.8484\n",
      "Epoch 77/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0384 - mean_iou: 0.8488\n",
      "Epoch 00077: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0384 - mean_iou: 0.8488 - val_loss: 0.1001 - val_mean_iou: 0.8493\n",
      "Epoch 78/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0382 - mean_iou: 0.8497\n",
      "Epoch 00078: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0379 - mean_iou: 0.8497 - val_loss: 0.1008 - val_mean_iou: 0.8501\n",
      "Epoch 79/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0408 - mean_iou: 0.8505\n",
      "Epoch 00079: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0406 - mean_iou: 0.8505 - val_loss: 0.0989 - val_mean_iou: 0.8509\n",
      "Epoch 80/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0386 - mean_iou: 0.8513\n",
      "Epoch 00080: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0388 - mean_iou: 0.8513 - val_loss: 0.0995 - val_mean_iou: 0.8517\n",
      "Epoch 81/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0368 - mean_iou: 0.8521\n",
      "Epoch 00081: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0365 - mean_iou: 0.8521 - val_loss: 0.1031 - val_mean_iou: 0.8525\n",
      "Epoch 82/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0362 - mean_iou: 0.8529\n",
      "Epoch 00082: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0362 - mean_iou: 0.8529 - val_loss: 0.1039 - val_mean_iou: 0.8533\n",
      "Epoch 83/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0355 - mean_iou: 0.8537\n",
      "Epoch 00083: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0356 - mean_iou: 0.8537 - val_loss: 0.1056 - val_mean_iou: 0.8542\n",
      "Epoch 84/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0365 - mean_iou: 0.8545\n",
      "Epoch 00084: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0365 - mean_iou: 0.8545 - val_loss: 0.1060 - val_mean_iou: 0.8549\n",
      "Epoch 85/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0352 - mean_iou: 0.8553\n",
      "Epoch 00085: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0353 - mean_iou: 0.8553 - val_loss: 0.1024 - val_mean_iou: 0.8557\n",
      "Epoch 86/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0353 - mean_iou: 0.8560\n",
      "Epoch 00086: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0353 - mean_iou: 0.8560 - val_loss: 0.1141 - val_mean_iou: 0.8564\n",
      "Epoch 87/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0346 - mean_iou: 0.8568\n",
      "Epoch 00087: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0351 - mean_iou: 0.8568 - val_loss: 0.1024 - val_mean_iou: 0.8572\n",
      "Epoch 88/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0360 - mean_iou: 0.8575\n",
      "Epoch 00088: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0358 - mean_iou: 0.8575 - val_loss: 0.1051 - val_mean_iou: 0.8579\n",
      "Epoch 89/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0350 - mean_iou: 0.8582\n",
      "Epoch 00089: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0351 - mean_iou: 0.8583 - val_loss: 0.1064 - val_mean_iou: 0.8586\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 90/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0338 - mean_iou: 0.8590\n",
      "Epoch 00090: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0335 - mean_iou: 0.8590 - val_loss: 0.1117 - val_mean_iou: 0.8594\n",
      "Epoch 91/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0340 - mean_iou: 0.8597\n",
      "Epoch 00091: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0338 - mean_iou: 0.8597 - val_loss: 0.1114 - val_mean_iou: 0.8601\n",
      "Epoch 92/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0325 - mean_iou: 0.8604\n",
      "Epoch 00092: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0328 - mean_iou: 0.8604 - val_loss: 0.1126 - val_mean_iou: 0.8608\n",
      "Epoch 93/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0321 - mean_iou: 0.8611\n",
      "Epoch 00093: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0323 - mean_iou: 0.8611 - val_loss: 0.1122 - val_mean_iou: 0.8615\n",
      "Epoch 94/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0316 - mean_iou: 0.8618\n",
      "Epoch 00094: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0315 - mean_iou: 0.8618 - val_loss: 0.1149 - val_mean_iou: 0.8622\n",
      "Epoch 95/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0314 - mean_iou: 0.8625\n",
      "Epoch 00095: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0313 - mean_iou: 0.8625 - val_loss: 0.1186 - val_mean_iou: 0.8629\n",
      "Epoch 96/100\n",
      "600/603 [============================>.] - ETA: 0s - loss: 0.0319 - mean_iou: 0.8632\n",
      "Epoch 00096: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0318 - mean_iou: 0.8632 - val_loss: 0.1107 - val_mean_iou: 0.8635\n",
      "Epoch 97/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0317 - mean_iou: 0.8639\n",
      "Epoch 00097: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0317 - mean_iou: 0.8639 - val_loss: 0.1124 - val_mean_iou: 0.8642\n",
      "Epoch 98/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0302 - mean_iou: 0.8645\n",
      "Epoch 00098: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0301 - mean_iou: 0.8645 - val_loss: 0.1154 - val_mean_iou: 0.8649\n",
      "Epoch 99/100\n",
      "592/603 [============================>.] - ETA: 0s - loss: 0.0320 - mean_iou: 0.8652\n",
      "Epoch 00099: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0321 - mean_iou: 0.8652 - val_loss: 0.1162 - val_mean_iou: 0.8655\n",
      "Epoch 100/100\n",
      "584/603 [============================>.] - ETA: 0s - loss: 0.0322 - mean_iou: 0.8658\n",
      "Epoch 00100: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0321 - mean_iou: 0.8658 - val_loss: 0.1181 - val_mean_iou: 0.8661\n"
     ]
    }
   ],
   "source": [
    "#earlystopper = EarlyStopping(patience=5, verbose=1)\n",
    "checkpointer = ModelCheckpoint('model-tmp-v3.h5', verbose=1, save_best_only=True)\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=100, verbose=1, callbacks=[checkpointer])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00001: val_loss did not improve\n",
      "\n",
      "Epoch 00002: val_loss did not improve\n",
      "\n",
      "Epoch 00003: val_loss did not improve\n",
      "\n",
      "Epoch 00004: val_loss did not improve\n",
      "\n",
      "Epoch 00005: val_loss did not improve\n",
      "\n",
      "Epoch 00006: val_loss did not improve\n",
      "\n",
      "Epoch 00007: val_loss did not improve\n",
      "\n",
      "Epoch 00008: val_loss did not improve\n",
      "\n",
      "Epoch 00009: val_loss did not improve\n",
      "\n",
      "Epoch 00010: val_loss did not improve\n",
      "\n",
      "Epoch 00011: val_loss did not improve\n",
      "\n",
      "Epoch 00012: val_loss did not improve\n",
      "\n",
      "Epoch 00013: val_loss did not improve\n",
      "\n",
      "Epoch 00014: val_loss did not improve\n",
      "\n",
      "Epoch 00015: val_loss did not improve\n",
      "\n",
      "Epoch 00016: val_loss did not improve\n",
      "\n",
      "Epoch 00017: val_loss did not improve\n",
      "\n",
      "Epoch 00018: val_loss did not improve\n",
      "\n",
      "Epoch 00019: val_loss did not improve\n",
      "\n",
      "Epoch 00020: val_loss did not improve\n",
      "\n",
      "Epoch 00021: val_loss did not improve\n",
      "\n",
      "Epoch 00022: val_loss did not improve\n",
      "\n",
      "Epoch 00023: val_loss did not improve\n",
      "\n",
      "Epoch 00024: val_loss did not improve\n",
      "\n",
      "Epoch 00025: val_loss did not improve\n",
      "\n",
      "Epoch 00026: val_loss did not improve\n",
      "\n",
      "Epoch 00027: val_loss did not improve\n",
      "\n",
      "Epoch 00028: val_loss did not improve\n",
      "\n",
      "Epoch 00029: val_loss did not improve\n",
      "\n",
      "Epoch 00030: val_loss did not improve\n",
      "Train on 603 samples, validate on 67 samples\n",
      "Epoch 1/1\n",
      "592/603 [============================>.] - ETA: 0s - loss: 0.0233 - mean_iou: 0.8819\n",
      "Epoch 00001: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0234 - mean_iou: 0.8819 - val_loss: 0.1406 - val_mean_iou: 0.8822\n"
     ]
    }
   ],
   "source": [
    "K.set_value(model.optimizer.lr, 0.001)\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=30, verbose=0, callbacks=[checkpointer])\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=1, verbose=1, callbacks=[checkpointer])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00001: val_loss did not improve\n",
      "\n",
      "Epoch 00002: val_loss did not improve\n",
      "\n",
      "Epoch 00003: val_loss did not improve\n",
      "\n",
      "Epoch 00004: val_loss did not improve\n",
      "\n",
      "Epoch 00005: val_loss did not improve\n",
      "\n",
      "Epoch 00006: val_loss did not improve\n",
      "\n",
      "Epoch 00007: val_loss did not improve\n",
      "\n",
      "Epoch 00008: val_loss did not improve\n",
      "\n",
      "Epoch 00009: val_loss did not improve\n",
      "\n",
      "Epoch 00010: val_loss did not improve\n",
      "\n",
      "Epoch 00011: val_loss did not improve\n",
      "\n",
      "Epoch 00012: val_loss did not improve\n",
      "\n",
      "Epoch 00013: val_loss did not improve\n",
      "\n",
      "Epoch 00014: val_loss did not improve\n",
      "\n",
      "Epoch 00015: val_loss did not improve\n",
      "\n",
      "Epoch 00016: val_loss did not improve\n",
      "\n",
      "Epoch 00017: val_loss did not improve\n",
      "\n",
      "Epoch 00018: val_loss did not improve\n",
      "\n",
      "Epoch 00019: val_loss did not improve\n",
      "\n",
      "Epoch 00020: val_loss did not improve\n",
      "\n",
      "Epoch 00021: val_loss did not improve\n",
      "\n",
      "Epoch 00022: val_loss did not improve\n",
      "\n",
      "Epoch 00023: val_loss did not improve\n",
      "\n",
      "Epoch 00024: val_loss did not improve\n",
      "\n",
      "Epoch 00025: val_loss did not improve\n",
      "\n",
      "Epoch 00026: val_loss did not improve\n",
      "\n",
      "Epoch 00027: val_loss did not improve\n",
      "\n",
      "Epoch 00028: val_loss did not improve\n",
      "\n",
      "Epoch 00029: val_loss did not improve\n",
      "\n",
      "Epoch 00030: val_loss did not improve\n",
      "Train on 603 samples, validate on 67 samples\n",
      "Epoch 1/1\n",
      "600/603 [============================>.] - ETA: 0s - loss: 0.0159 - mean_iou: 0.8953\n",
      "Epoch 00001: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0159 - mean_iou: 0.8953 - val_loss: 0.1949 - val_mean_iou: 0.8955\n"
     ]
    }
   ],
   "source": [
    "K.set_value(model.optimizer.lr, 0.0001)\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=30, verbose=0, callbacks=[checkpointer])\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=1, verbose=1, callbacks=[checkpointer])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "Epoch 00001: val_loss did not improve\n",
      "\n",
      "Epoch 00002: val_loss did not improve\n",
      "\n",
      "Epoch 00003: val_loss did not improve\n",
      "\n",
      "Epoch 00004: val_loss did not improve\n",
      "\n",
      "Epoch 00005: val_loss did not improve\n",
      "\n",
      "Epoch 00006: val_loss did not improve\n",
      "\n",
      "Epoch 00007: val_loss did not improve\n",
      "\n",
      "Epoch 00008: val_loss did not improve\n",
      "\n",
      "Epoch 00009: val_loss did not improve\n",
      "\n",
      "Epoch 00010: val_loss did not improve\n",
      "\n",
      "Epoch 00011: val_loss did not improve\n",
      "\n",
      "Epoch 00012: val_loss did not improve\n",
      "\n",
      "Epoch 00013: val_loss did not improve\n",
      "\n",
      "Epoch 00014: val_loss did not improve\n",
      "\n",
      "Epoch 00015: val_loss did not improve\n",
      "\n",
      "Epoch 00016: val_loss did not improve\n",
      "\n",
      "Epoch 00017: val_loss did not improve\n",
      "\n",
      "Epoch 00018: val_loss did not improve\n",
      "\n",
      "Epoch 00019: val_loss did not improve\n",
      "\n",
      "Epoch 00020: val_loss did not improve\n",
      "\n",
      "Epoch 00021: val_loss did not improve\n",
      "\n",
      "Epoch 00022: val_loss did not improve\n",
      "\n",
      "Epoch 00023: val_loss did not improve\n",
      "\n",
      "Epoch 00024: val_loss did not improve\n",
      "\n",
      "Epoch 00025: val_loss did not improve\n",
      "\n",
      "Epoch 00026: val_loss did not improve\n",
      "\n",
      "Epoch 00027: val_loss did not improve\n",
      "\n",
      "Epoch 00028: val_loss did not improve\n",
      "\n",
      "Epoch 00029: val_loss did not improve\n",
      "\n",
      "Epoch 00030: val_loss did not improve\n",
      "Train on 603 samples, validate on 67 samples\n",
      "Epoch 1/1\n",
      "576/603 [===========================>..] - ETA: 0s - loss: 0.0151 - mean_iou: 0.9051\n",
      "Epoch 00001: val_loss did not improve\n",
      "603/603 [==============================] - 1s 2ms/step - loss: 0.0151 - mean_iou: 0.9051 - val_loss: 0.2042 - val_mean_iou: 0.9053\n"
     ]
    }
   ],
   "source": [
    "K.set_value(model.optimizer.lr, 0.00001)\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=30, verbose=0, callbacks=[checkpointer])\n",
    "results = model.fit(X_train, Y_train, validation_split=0.1, batch_size=8, epochs=1, verbose=1, callbacks=[checkpointer])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "603/603 [==============================] - 0s 739us/step\n",
      "67/67 [==============================] - 0s 370us/step\n",
      "65/65 [==============================] - 0s 620us/step\n"
     ]
    }
   ],
   "source": [
    "# Predict on train, val and test\n",
    "model = load_model('model-tmp-v3.h5', custom_objects={'mean_iou': mean_iou})\n",
    "preds_train = model.predict(X_train[:int(X_train.shape[0]*0.9)], verbose=1)\n",
    "preds_val = model.predict(X_train[int(X_train.shape[0]*0.9):], verbose=1)\n",
    "preds_test = model.predict(X_test, verbose=1)\n",
    "\n",
    "# Threshold predictions\n",
    "preds_train_t = (preds_train > 0.5).astype(np.uint8)\n",
    "preds_val_t = (preds_val > 0.5).astype(np.uint8)\n",
    "preds_test_t = (preds_test > 0.5).astype(np.uint8)\n",
    "\n",
    "# Create list of upsampled test masks\n",
    "preds_test_upsampled = []\n",
    "for i in range(len(preds_test)):\n",
    "    preds_test_upsampled.append(resize(np.squeeze(preds_test[i]), \n",
    "                                       (sizes_test[i][0], sizes_test[i][1]), \n",
    "                                       mode='constant', preserve_range=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nPzOVDgGibUkrAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb18db2410>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb18a9d990>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUQAAAEYCAYAAAAkpo9KAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3X/QZFWd3/H3h0ElgyLiGDIOGGbLYTezBJWagJapSATD\nYCyoVCwy6LqgmKlUieuquwvElBpTVmHc6GKFsDuLLGixIrLuMmVmfRZZKSupiAwrNcIg+Cz+YPjh\ngCK7lYnKzHzzx70P0/R0973dfX+ce/vzqup6nr59+97Tp2+f+73nnHuOIgIzM4Mj2k6AmVkqXCCa\nmeVcIJqZ5VwgmpnlXCCameVcIJqZ5VwgmlknSbpO0l5J9455XZI+I2lZ0i5JpxVts7YCUdJmSQ/k\nibm8rv2Y2cK6Htg84fVzgQ35YytwTdEGaykQJa0Crs4TtBG4UNLGOvZlZospIr4B/HTCKucDn4vM\nN4FjJa2dtM0jq0zggNOB5Yh4CEDSTXnido9a+fl6QRzF0c8+P/nUfYU7eHDX6koSamkadQy0+Z0X\nHZNNpG04DdPss8xvatptFvl7nnoyIl4GcM6/PDp+8tMDpd97965f3Af8fGDRtojYNmUS1gEPDzzf\nky97bNwb6ioQRyXkjMEVJG0lC2M5itWcobOefW1p6Z7CHZzz8ldXkU5L1KhjoM3vvOiYbCJtw2mY\nZp9lflPTbrPI1+KWH678/5OfHuBbS68o/d5Va7/384jYVFliSqqrQCyUl/bbAI7RcQGw9Gi5L21w\n3ZQKxnHpTymNqZvmGLBiqeRnAAc52PRuHwFOHHh+Qr5srLoaVaZOiJn1WXAgDpZ+VGQ78Jt5a/Nr\ngacjYuzlMtQXId4FbJC0nqwg3AK8rehNK5FUKme1qqQYzaamzHfeZj729dhsShYhVjuylqQvAGcC\nayTtAT4CPA8gIv4Q2AG8GVgG9gHvLNpmLQViROyXdCmwBKwCrouI++rYl5l1Q9WXzBFxYcHrAbxn\nmm3WVocYETvISmizQtNEYClEim3o8hVGEBzowNirrTWqmNliqfqSuQ5JFohlooWUzpauV0rD8PeQ\n0jGSurrzKoADLhDNzDKOEOfkM3waUu1fWRSZu3W/WFN5E8AzrkM0M8sbVRwhlnPyqfsm3lrU9TN8\nmfSnWP9VJgKrOp3D23P97HyG6+PLfl+VR9cBB9IvD9MoEIukfukzrhFoUnp9uVfOOS9/9dyFovNy\n+oJw1PN58i/rmJ2+ThSIZtZ14gBqOxGFOlUgtn2mL9p/ahFIqo0hbWn7+Om6efIvgIO+ZDYzyzhC\n7DhX6BdrIuoat+1x+04tMk4tPaOUOdbnrUN0gWhmljsYLhBLeXDX6kpaE5vSZlRU1qS8LJv+1Ie8\nKqrLTSHdVXwPfeAI0cwsF4gDHZj1OKkCMaUze1tSjBYmfS8ppndFymkblEKkWOa3N29dqC+Zzczw\nJfNcFiFSrDsaKFMnO21k0pWIKyVdqhuHOn974kD4ktnMLL91zwXiXByRWFPKRkSzHJNFUdeiHOe+\nZDYzAyJ8yWwt60tdbJ3R2zTbH7XuNPtclEhwnIOOEM3MVlqZHSEmp4p6nDbO9Ita/zRL9Nb3PCmj\nzryYbZu+ZDYzA9zK3At11cFNM/R+Fft2xFS9VPN03IjX06a3js93wHeqmJn5XubkzHPXRlVny8E0\nVFXH43qzZjmfZ5NNQ5p+cZN+Cs2s8wL5krms4WlI+3YW7no/wL5L7Xibtq9jV44vN6qYmQER9Lvb\njaQTgc8Bx5NVEWyLiKskHQd8ETgJ+AFwQUQ8Nc22q6wX68rZsy+K8nuW1s6+f4d9/3wZ9f5Olf3A\nByPibyS9CLhb0m3AxcDtEXGlpMuBy4HLZtlBkw0GdeyjzIHel0aROn/UZSeR6oqq0j9pO6kdT0HP\nI8SIeAx4LP//7yXdD6wDzgfOzFe7AbiDGQtEM+uPhel2I+kk4DXAncDxeWEJ8DjZJfWo92wFtgK8\nYl03qjKnrexuYnCF1CKBJnT1M1d9HHRp8I5AnZhCYO4iW9ILgT8Dfjsi/m7wtYgIsmj5MBGxLSI2\nRcSml7101bzJMLPEHeCI0o+2zBWaSXoeWWF4Y0R8OV/8Y0lrI+IxSWuBvfMmsi1NnHmnjTJniY6a\nGBiiqmilrYmsqm4MaksK6VzJy1VrDy0L4GAH6hBnTqEkAZ8F7o+ITw28tB24KP//IuDW2ZNnZv0g\nDkzxaMs8EeLrgXcA35G0cnr9j8CVwM2SLgF+CFww6w6qONv1Zfj2KiPD4de7kgeLIoWBGKrWlQhx\nnlbm/wVji/KzZt2umfWT51Qp6cFdqxuZlhPKR0VV1h+mNKDspHXrjMjnye+6I9lUWmlTjPRmyftD\n6y4/uyxClUeIkjYDVwGrgGsj4sqh119B1vXv2HydyyNix6RtJlEgmln/VdkxW9Iq4GrgTcAe4C5J\n2yNi98Bq/wm4OSKukbQR2EF2B91YC1cgpngW7iPn8+Gq6DHQlFEDzc6T3mzE7EovmU8HliPiIQBJ\nN5HdFDJYIAZwTP7/i4FHiza6cAWimbVh6jlV1kjaOfB8W0RsG3i+Dnh44Pke4IyhbXwU+CtJ7wWO\nBs4u2qkLRKtViq3709QZpjNJU7dlrcxTRYhPRsSmOXd7IXB9RPw3Sa8DPi/plIg4OO4NLhDNrBEV\n34HyCHDiwPMT8mWDLgE2A0TE/5F0FLCGCTeLuEAcY9KwU104w08zbFbqn6eOKRxstLryqIZ7me8C\nNkhaT1YQbgHeNrTOj8i6AF4v6Z8ARwFPTNqoC0Qza0SVI2ZHxH5JlwJLZF1qrouI+yR9DNgZEduB\nDwJ/LOn9ZFftF+fjK4zlAnGC1COnIl2pp0tJ17/zsppu6c5GzK62Y3bep3DH0LIPD/y/m+yOutJc\nIJpZI7ow/JcLxI5ru09bVfuf9Q6XqixaZDj8vO7PH4hnIv1h/lwgmlntZuh204qkC8Qy9VApndmb\nbJVu60w/Lg1V1Rmm9H32ybzH5vzfS/X3Mtch6QLRzPqj77Pu1aatkVpmldK4g4NpcbTVD/PUq6bS\n2l9HK3MdkiwQuyKVg836adzxVeYEO+uxWedJ1JfMZmZ0Z9Y9F4jW6VsU+2TaqG6eqpg2vlvXIZqZ\n4W43c+nSBNxtqWLyrHkbg7o2gETbnb+r1pXIcIXrEM3MAMJ1iHMbPpu1NYl5yprsyjNqX12e5jWV\ntNV5JZTKZ6xhCoFaJF0gmll/OEIs6eRT97G0VDyJTSpnu2l1Nd3DykSKXTAqrUVRWhWfr8nbG1P7\nPtyoYmY2wAXilFK4Dc8WT5nILYVjs8u/C3fMNjMb4EaVnutyCyu4v2fqn7vK46f1YzR8yWxmBixQ\no4qkVcBO4JGIeEs+LeBNwEuBu4F3RMQv591P1ao8Yw5HWqlHhsMWLVKc53O2HmlNKaWh6bpQIFZx\nL837gPsHnn8C+HREvBJ4imyyaDNbYCuNKmUfbZkrQpR0AvCvgY8DH5Ak4I0cmjD6BuCjwDWTtvPg\nrtW1n6GaiH7qGLizycgj1SgnBUVRdJX3R6fU57FK0YEIcd5L5j8Afg94Uf78pcDPImJ//nwPsG7U\nGyVtBbYCHMXqOZNhZqnrdSuzpLcAeyPibklnTvv+iNgGbAM4RsfFrOmoWhN1KimevW20OqO1LkXk\nS48W30k2SSxAK/PrgfMkvRk4CjgGuAo4VtKReZR4AvDI/Mk0s24TBw72ePiviLgCuAIgjxB/JyLe\nLulLwFvJWpovAm6tIJ0z63o01tWW66rUMWnWqPrAuvI39eOv6HOPmmp21rzqQh1iHUX2ZWQNLMtk\ndYqfrWEfZtYhK/0Qe93KvCIi7gDuyP9/CDi9iu3a4pkUUVUdLbcVdacQ7U8TGVYisnrE1PlOFTNr\nRK9bmW1208xFYmnpyv3rdc4fM0ukHnSjDrH3BeK0t6U1dUCXmR7BmjPv5fikibjK7nuW9xZJpYDG\nw3+ZmR3iOsQOSedMOlpfps5MUZVdS6qUUlqq4EtmMzOy6NAF4pw8RNd4ZT5HExMn9dW4/C0a3KGO\nPG3jmK1jSDjXIZqZ5VyHOKNFmLg7BYtSL1n0Od0N6nBVDKK7au1zl/uS2cyMbIBYF4hT8lm6XX2p\nX11RRX1fkwMLd+X4H5dvRenvwBVzLYM7mJk9V97KXPZRhqTNkh6QtCzp8jHrXCBpt6T7JP1p0TaT\nihAtDV2PFPvWup5ay/XM760wRMwnt7saeBPZyPx3SdoeEbsH1tlANkTh6yPiKUn/sGi7jhDNrBEV\nR4inA8sR8VA+q+dNwPlD6/x74OqIeCrbf+wt2mgSEeLJp+5jaam4/qRrZ/ZplB2Oqc95UKTOIaum\nHbShq9/H8Ocb/hx11mNW3O1mHfDwwPM9wBlD65wMIOl/A6uAj0bEVydtNIkC0cz6bYbRbtZI2jnw\nfFs+D9M0jgQ2AGeSTWfyDUn/NCJ+NukNNmDUGTKFKKCKu3OsWNk8K7qDZZ7vK3UzpTOA6QrEJyNi\n04TXHwFOHHg+av6mPcCdEfEM8H1JD5IVkHeN26jrEM2sERHlHyXcBWyQtF7S84EtwPahdf6CLDpE\n0hqyS+iHJm20ExGipwSdT5cixSrTOi6KK3MVMG99ZSutuHMok99F6Sn87iqsQ4yI/ZIuBZbI6gev\ni4j7JH0M2BkR2/PX/pWk3cAB4Hcj4ieTttuJAtHMuk7EwWrvVImIHcCOoWUfHvg/gA/kj1I6USDW\nOQy6HS61uqyy6Wki3XUeN+NagFdMc4dI1SOyz523Hv7LzGxAB+7d62SBuPToPTOfsRY5Mkwt8puk\nyrT29TtvYsrWad5fnM+OEM3MMo4Q0zFLlNCliMr6b5aor4nouPQ+XCCamTFLx+xWJFEgPrhr9XPO\neinU+Tg6tFFS6NNZ5/3HdR73nkKgg1wQ9k9XCo4UCttauUA0M8v1/ZJZ0rHAtcApZOX/u4AHgC8C\nJwE/AC5YGY8sZY4M+68r33HZSHGaSLKOwSimjWTVgQhx3sEdrgK+GhG/BrwKuB+4HLg9IjYAt+fP\nzWyRxZSPlswcIUp6MfAvgIsB8lFrfynpfPIRJoAbgDuAy+ZJ5LAqz/RdiRqsGWWjnq4fN1Wkf7o6\nT3XiknmeCHE98ATwJ5K+LelaSUcDx0fEY/k6jwPHj3qzpK2Sdkra+Qy/mCMZZtYJfY4Q8/eeBrw3\nIu6UdBVDl8cREdLomoN89NttAMfouNqzIIUz+qQzaQrps/KaGKJrOAKresCGxvW8DnEPsCci7syf\n30JWQP5Y0lqA/G/hxC5mtgD6HCFGxOOSHpb0qxHxAHAWsDt/XARcmf+9ddptF50ZLT19mfpz2r6A\nTUaKqRqfvuVD/y7InSrvBW7Mh/B+CHgnWdR5s6RLgB8CF8y5DzPrgS50u5mrQIyIe4BRE8GcNc92\nV6R+ZqxT2ekwzTqjAwWiJ5kyM8v51r2O6Wq9alfTnbIy9Z0p5XfvL5nNzEpbgEYV64E6JvFqQp31\nrF0aeaYT31fL3WnKcoFoZo3QwbZTUMwFYiKmrWOrok5uXPSTen1fnZPEDyuKFFPNoyQ5QjQzy7lA\ntLrVOfFQapFim/V5qeRBVyncymxmdohbmW1Ql1ouq9ZGhJVahLvwHCGamWV8ydyCro66kmq6yhoX\n/Xb9c1mFXCCamQFuVLFxuhI1zZLOJscFbLIudlH7IVZ6n7QLxOZM++Mos36bB3udDTDjCpau/bhT\naKTqaxVBmTyd+iThAtHMLONL5gR1rctLE1FQ16OaKtJfVZTcVrRddWPiPMfdyntWrZ36ra1buALR\nzFriCLE5KdQn1anrUVzKBo+ZLh0/06S19XpitzKbmQ1wgZiespGkI7J0FEU3fW3pHdal6HUkF4hm\nZiB8yTy1Js/0fa9z7ItR9XtFx0Xr9WUJmzZvKv2dVFwgStoMXAWsAq6NiCvHrPdvgVuAfxYROydt\n09OQmln94tCYiGUeRSStAq4GzgU2AhdK2jhivRcB7wPuLJPMJCLEk0/dx9LS4WegWc70HmSgv+q6\nde+cl7+60m31UZk8OvyzLz/3abUR4unAckQ8BCDpJuB8YPfQev8F+ATwu2U26gjRzJoRUzyKrQMe\nHni+J1/2LEmnASdGxP8sm8QkIsQi80SK8+yzr2d7q94iHCvzfsYpG1XWSBqs79sWEdtK70s6AvgU\ncPE0O+1EgWhmHRfAdNOQPhkRmya8/ghw4sDzE/JlK14EnALcIQngHwHbJZ03qWHFBeIEbq1MSxP3\nc7t/an0q7nZzF7BB0nqygnAL8LaVFyPiaWDNs/uW7gB+x63MZpaGCusQI2I/cCmwBNwP3BwR90n6\nmKTzZk3iXBGipPcD7yb7CN8B3gmsBW4CXgrcDbwjIn45z37q5P6I/TNPFJdyL4UqjtU2P0fVHbMj\nYgewY2jZh8ese2aZbc4cIUpaB/wWsCkiTiHrHLmFrIn70xHxSuAp4JJZ92FmPVJtK3Mt5q1DPBL4\nB5KeAVYDjwFv5NC1/A3AR4Fr5tlJ22fnvtUlThthdOVzV5nOlD/zqLSlVPc5cjzElgu6smaOECPi\nEeD3gR+RFYRPk10i/yy/vocRfYNWSNoqaaeknU/85MCsyTCzDtCUj7bMHCFKeglZz/D1wM+ALwGb\ny74/71O0DWDTq44aee5I+Sy9SFKNkFNLT5s6kRcdiBDnuWQ+G/h+RDwBIOnLwOuBYyUdmUeJw32D\nzGxB9X20mx8Br5W0Gvh/wFnATuDrwFvJWpovAm4t2tCDu1Z34wzXcfO2pKcaKVpHdKBAnKcO8U6y\nIXX+hqzLzRFkl8CXAR+QtEzW9eazFaTTzLqu763MEfER4CNDix8iG4miEVX1F3N/xDRVORLNikon\nX7fDHMrDgdFuPKdKvcYd1PNe1rlgLNb0pfPwd1Lnfl0tUCMXiGZmGUeIHTZ4qVZHtJDirWGp60se\nLWwU6gLRzCzjCLHj6jqDj6qfbCJqqKp+tM+RTRP1k+OeT7vvwfcn/5105NY9F4hm1gwXiO1Z2Hqa\njhsVvfo7fK5JVxgrUsszz8tsZjbIBWJ93F9wdtPmXRPRRtvfYwoRVd+vahTpl4idLRDNrEPcqGKp\nSyESKRMZVh05pfC5ZzFNFJ1itOk6RDOznKabhrQVLhCtFW3XGc4q9dbcpDlCNDPDo900xa3N/Tdv\nFFZFVFfX6Ep1meeul7r24QjRzAx3zG5ck2PmWTc0GdWV2WYTVzNtTDdaet/uh2hmlnGEWNLJp+5j\naemeSs5ujgz7ZRG/z3HRZup5MTFKdsdsM7ND3A/RRho1cVLqZ/+m9Tk/iuoSu/7ZVz7XqrVDLzhC\nNDPLuA6xA9o6S3c9CphXCp+/rdbmFQvVhzZwK3OXLT1aTSOPWZEmJjGrc1/j97H8nOWOEM3MVrhA\nLOfBXauTjMbcydtWdOXytih9dR7Tk7bpO1XMzFZEuA7R6lMUCTiqtTIaGdQh5wjRzGxFBwrEI4pW\nkHSdpL2S7h1Ydpyk2yR9L//7kny5JH1G0rKkXZJOqzPxTVl69J7k646GdS291n+K8o+2FBaIwPXA\n5qFllwO3R8QG4Pb8OcC5wIb8sRW4pppkmlmnBXAwyj9aUnjJHBHfkHTS0OLzgTPz/28A7gAuy5d/\nLiIC+KakYyWtjYjHqkpwm7rW6ty19FatKy3DXTH3cdSHS+Yxjh8o5B4Hjs//Xwc8PLDennzZYSRt\nlbRT0s5n+MWMyTCzrujCJfPcjSoREdL0HyEitgHbAI7RcSPf38QACLNEEW1GXn2NdtoY7GKWfbQR\ndaYwCEQl+6q4242kzcBVwCrg2oi4cuj1DwDvBvYDTwDviogfTtrmrBHijyWtzXe6FtibL38EOHFg\nvRPyZWa2yCIb/qvso4ikVcDVZO0WG4ELJW0cWu3bwKaIOBW4BfivRdudNULcDlwEXJn/vXVg+aWS\nbgLOAJ6epf6waxP62OxmicyHFR0PXT1e6rjrZNygs3VHodmdKpVGiKcDyxHxEEBe5pwP7F5ZISK+\nPrD+N4HfKNpoYYEo6QtkDShrJO0BPkJWEN4s6RLgh8AF+eo7gDeT3dW9D3hn0fbNbEFMN0DsGkk7\nB55vy6vZVoxqrzhjwvYuAf6yaKdlWpkvHPPSWSPWDeA9Rduc1+DZrKozmFsku6uNK4dxx0sVaZj2\nGBwcmWnWdI17X5V5O2WE+GREbJp7p4Ck3wA2AW8oWtd3qphZ/aqfU6VUe4Wks4EPAW+IiMLuLC4Q\nhzhSTE/K30Wq9ZNNTFQ/ncoHd7gL2CBpPVlBuAV42+AKkl4D/BGwOSL2Hr6Jw7lANLNGVNm/MCL2\nS7oUWCLrdnNdRNwn6WPAzojYDnwSeCHwJUkAP4qI8yZt1wXiGINn2JQjlDKqqufyxFjNaPIqpew+\nKhlBvuJ+iBGxg6whd3DZhwf+P3vabbpANLP6hachndnwmajtCG3RI6FJ/UIXPW9S0JnvwAPEmpnl\n0i8P0ywQh/s+jYoYO3NW7LC2I/NF17djvOI7VWqRZIFoZj3kArGck0/dx9LS4dHIuF7yfTtz9lHZ\ne2Pd73NBBNPeuteKJApEWxw+yaWlqROSCF8ym5k9ywWidYE7XFsj37kLRDMzXIdo6SsaiLdIKpFk\nKumoyzT1eynnhesQzcxWuECczzRnuxQm4klBmVbDeVsUJ+VpHa2VdQ7G2jdFA7oWfT/z/OYmv7fy\n4b9qkXSBaGY9EbhAnJXP/OO10ZE5le8jlXQsupkngXOjiplZRgfTLxGTKBAf3LW6skFLF8U80VId\nw6s10TLtaWjrV0Uer2xj1dqBhQEc9CWzmRluVDEbUEXEMfw85UixyjRWUW+cxJWUC0Qzs5wLRGvL\nOS9/9dz99lKIwJKIbKYwLpqF+fOz00OluQ7RzGxFQLiV2VpUNiKpMxKsc9sp1SW20S+0zruCauFL\nZjMzfMlsiy2FqC01VUe0VUaKHg8xc0TRCpKuk7RX0r0Dyz4p6buSdkn6c0nHDrx2haRlSQ9IOqeu\nhJtZx0SUf7SkTIR4PfDfgc8NLLsNuCIi9kv6BHAFcJmkjcAW4NeBlwNfk3RyRByoNtlp8Ug79X3W\nUa3lw1KqSyyjrkixzDrtjRjUjY7ZhRFiRHwD+OnQsr+KiP35028CJ+T/nw/cFBG/iIjvA8vA6RWm\n18y6KICDB8s/WlJFHeK7gC/m/68jKyBX7MmXHUbSVmArwFGsriAZo9UZPXSyP1gPpRAZdqWPYJV3\nzqwo/Zk7ECHOVSBK+hCwH7hx2vdGxDZgG8AxOi79nDKz+fS5QJR0MfAW4KyIZz/pI8CJA6udkC/r\nldSjgL7pSvS1qMZHncsD/0d/u91I2gz8HvCGiNg38NJ24E8lfYqsUWUD8K25U5kI/yCrM8tQ9ilc\nGhdx4T1GQPThThVJXwDOBNZI2gN8hKxV+QXAbZIAvhkR/yEi7pN0M7Cb7FL6PX1vYTazkvoQIUbE\nhSMWf3bC+h8HPj5PooZNahgpOiN3Iaqww3WtK82wUcdlyp+lkd9Pn+sQzcxKi2i1O01ZSReIs0zQ\nnULdzSxn1a5HRGWl8P1MUnWklPr32Wind0eIZmaZcIQ4m3miiDbPyPNEhuOepx5hLIpFieDr041b\n95IsEM2sZzz8V/XaPktXWU+Zel1aXVKq67XmBBAH0u+BVzi4g5nZ3CKfQqDsowRJm/NhBpclXT7i\n9RdI+mL++p2STiraZqcixFS5Xml6jhQXT1R4ySxpFXA18CayQWTukrQ9InYPrHYJ8FREvFLSFuAT\nwL+btF1HiGbWjGojxNOB5Yh4KCJ+CdxENvzgoPOBG/L/bwHOUn5r3TiKBFp+JD0B/F/gybbTMsYa\n0kyb0zW9VNPWx3T944h4GYCkr+bbKuso4OcDz7flI2SRb++twOaIeHf+/B3AGRFx6cA69+br7Mmf\n/22+ztjPk8Qlc0S8TNLOiNjUdlpGSTVtTtf0Uk1b39MVEZurSE/dfMlsZl1UZqjBZ9eRdCTwYuAn\nkzbqAtHMuuguYIOk9ZKeTzaX0/ahdbYDF+X/vxX46yioI0zikjm3rXiV1qSaNqdreqmmzemaQj7B\n3aXAErAKuC4ffvBjwM6I2E42KtfnJS2TzQu1pWi7STSqmJmlwJfMZmY5F4hmZrkkCsSiW3AaTMeJ\nkr4uabek+yS9L19+nKTbJH0v//uSltK3StK3JX0lf74+vyVpOb9F6fktpetYSbdI+q6k+yW9LoU8\nk/T+/Hu8V9IXJB3VVp5Juk7S3rxv3MqykXmkzGfyNO6SdFrD6fpk/l3ukvTnko4deO2KPF0PSDqn\nrnS1pfUCceAWnHOBjcCFkja2lJz9wAcjYiPwWuA9eVouB26PiA3A7fnzNrwPuH/g+SeAT0fEK4Gn\nyG5VasNVwFcj4teAV5GlsdU8k7QO+C1gU0ScQlbxvnL7Vht5dj0w3BdvXB6dSzZB2wayucuvaThd\ntwGnRMSpwINkcyiR/xa2AL+ev+d/5L/f/oiIVh/A64ClgedXAFe0na48LbeS3Sv5ALA2X7YWeKCF\ntJxA9qN5I/AVQGR3EBw5Kh8bTNeLge+TN9ANLG81z4B1wMPAcWS9Kb4CnNNmngEnAfcW5RHwR8CF\no9ZrIl1Dr/0b4Mb8/+f8NslaeF/X9DFX56P1CJFDB+6KPfmyVuUjY7wGuBM4PiIey196HDi+hST9\nAdnUrys3er4U+FlE7M+ft5Vv64EngD/JL+evlXQ0LedZRDwC/D7wI+Ax4GngbtLIsxXj8iil38S7\ngL/M/08pXbVIoUBMjqQXAn8G/HZE/N3ga5GdGhvtqyTpLcDeiLi7yf2WdCRwGnBNRLyG7J7051we\nt5RnLyG7uX892RzhR3P4pWEy2sijIpI+RFaNdGPbaWlKCgVimVtwGiPpeWSF4Y0R8eV88Y8lrc1f\nXwvsbThZrwfOk/QDslE93khWb3dsfksStJdve4A9EXFn/vwWsgKy7Tw7G/h+RDwREc8AXybLxxTy\nbMW4PGr9NyHpYuAtwNvzwjqJdNUthQKxzC04jciHBvoscH9EfGrgpcFbgC4iq1tsTERcEREnRMRJ\nZPnz1xGAXu09AAABCElEQVTxduDrZLcktZKuPG2PAw9L+tV80VnAblrOM7JL5ddKWp1/ryvpaj3P\nBozLo+3Ab+atza8Fnh64tK6dpM1k1TPnRcS+ofRuUTbw6nqyRp9vNZWuRrRdiZmffN5M1pr1t8CH\nWkzHPye7bNkF3JM/3kxWX3c78D3ga8BxLabxTOAr+f+/QnZALgNfAl7QUppeDezM8+0vgJekkGfA\nfwa+C9wLfB54QVt5BnyBrC7zGbKo+pJxeUTWYHZ1/nv4DllLeZPpWiarK1z5DfzhwPofytP1AHBu\nG8dbnQ/fumdmlkvhktnMLAkuEM3Mci4QzcxyLhDNzHIuEM3Mci4QzcxyLhDNzHL/H0FKMeU3MuGd\nAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb189d26d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Perform a sanity check on some random training samples\n",
    "ix = random.randint(0, len(preds_train_t))\n",
    "imshow(X_train[ix])\n",
    "plt.show()\n",
    "imshow(np.squeeze(Y_train[ix]))\n",
    "plt.show()\n",
    "imshow(np.squeeze(preds_train_t[ix]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb18988890>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAASEAAAEYCAYAAAATaEB+AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb187ddc90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUQAAAEYCAYAAAAkpo9KAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nZokDoplZ4oBoZpY4IJqZJf8fVOXUDqHfZHYAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7fbb186c0890>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Perform a sanity check on some random validation samples\n",
    "ix = random.randint(0, len(preds_val_t))\n",
    "imshow(X_train[int(X_train.shape[0]*0.9):][ix])\n",
    "plt.show()\n",
    "imshow(np.squeeze(Y_train[int(Y_train.shape[0]*0.9):][ix]))\n",
    "plt.show()\n",
    "imshow(np.squeeze(preds_val_t[ix]))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run-length encoding stolen from https://www.kaggle.com/rakhlin/fast-run-length-encoding-python\n",
    "def rle_encoding(x):\n",
    "    dots = np.where(x.T.flatten() == 1)[0]\n",
    "    run_lengths = []\n",
    "    prev = -2\n",
    "    for b in dots:\n",
    "        if (b>prev+1): run_lengths.extend((b + 1, 0))\n",
    "        run_lengths[-1] += 1\n",
    "        prev = b\n",
    "    return run_lengths\n",
    "\n",
    "def prob_to_rles(x, cutoff=0.5):\n",
    "    lab_img = label(x > cutoff)\n",
    "    for i in range(1, lab_img.max() + 1):\n",
    "        yield rle_encoding(lab_img == i)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "new_test_ids = []\n",
    "rles = []\n",
    "for n, id_ in enumerate(test_ids):\n",
    "    rle = list(prob_to_rles(preds_test_upsampled[n]))\n",
    "    rles.extend(rle)\n",
    "    new_test_ids.extend([id_] * len(rle))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create submission DataFrame\n",
    "sub = pd.DataFrame()\n",
    "sub['ImageId'] = new_test_ids\n",
    "sub['EncodedPixels'] = pd.Series(rles).apply(lambda x: ' '.join(str(y) for y in x))\n",
    "sub.to_csv('sub-tmp-v3.csv', index=False)"
   ]
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.14"
  },
  "nav_menu": {},
  "nbpresent": {
   "slides": {
    "28b43202-5690-4169-9aca-6b9dabfeb3ec": {
     "id": "28b43202-5690-4169-9aca-6b9dabfeb3ec",
     "prev": null,
     "regions": {
      "3bba644a-cf4d-4a49-9fbd-e2554428cf9f": {
       "attrs": {
        "height": 0.8,
        "width": 0.8,
        "x": 0.1,
        "y": 0.1
       },
       "content": {
        "cell": "f3d3a388-7e2a-4151-9b50-c20498fceacc",
        "part": "whole"
       },
       "id": "3bba644a-cf4d-4a49-9fbd-e2554428cf9f"
      }
     }
    },
    "8104def2-4b68-44a0-8f1b-b03bf3b2a079": {
     "id": "8104def2-4b68-44a0-8f1b-b03bf3b2a079",
     "prev": "28b43202-5690-4169-9aca-6b9dabfeb3ec",
     "regions": {
      "7dded777-1ddf-4100-99ae-25cf1c15b575": {
       "attrs": {
        "height": 0.8,
        "width": 0.8,
        "x": 0.1,
        "y": 0.1
       },
       "content": {
        "cell": "fe47bd48-3414-4657-92e7-8b8d6cb0df00",
        "part": "whole"
       },
       "id": "7dded777-1ddf-4100-99ae-25cf1c15b575"
      }
     }
    }
   },
   "themes": {}
  },
  "toc": {
   "navigate_menu": true,
   "number_sections": true,
   "sideBar": true,
   "threshold": 6,
   "toc_cell": false,
   "toc_section_display": "block",
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}