Model_02.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import numpy as np # linear algebra\n",
    "import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "\n",
    "from IPython.display import SVG\n",
    "from keras.utils.vis_utils import model_to_dot\n",
    "\n",
    "%matplotlib inline\n",
    "np.random.seed(666)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Load data\n",
    "train = pd.read_json(\"input/train.json\")\n",
    "test = pd.read_json(\"input/test.json\")\n",
    "\n",
    "# replace NAs with 0 in inc_angle column.\n",
    "train.inc_angle = train.inc_angle.replace('na', 0)\n",
    "train.inc_angle = train.inc_angle.astype(float).fillna(0.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_stats(train,label=1):\n",
    "    train['max'+str(label)] = [np.max(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['maxpos'+str(label)] = [np.argmax(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['min'+str(label)] = [np.min(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['minpos'+str(label)] = [np.argmin(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['med'+str(label)] = [np.median(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['std'+str(label)] = [np.std(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['mean'+str(label)] = [np.mean(np.array(x)) for x in train['band_'+str(label)] ]\n",
    "    train['p25_'+str(label)] = [np.sort(np.array(x))[int(0.25*75*75)] for x in train['band_'+str(label)] ]\n",
    "    train['p75_'+str(label)] = [np.sort(np.array(x))[int(0.75*75*75)] for x in train['band_'+str(label)] ]\n",
    "    train['mid50_'+str(label)] = train['p75_'+str(label)]-train['p25_'+str(label)]\n",
    "\n",
    "    return train\n",
    "train = get_stats(train,1)\n",
    "train = get_stats(train,2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_var(name,nbins=50):\n",
    "    minval = train[name].min()\n",
    "    maxval = train[name].max()\n",
    "    plt.hist(train.loc[train.is_iceberg==1,name],range=[minval,maxval],\n",
    "             bins=nbins,color='b',alpha=0.5,label='Boat')\n",
    "    plt.hist(train.loc[train.is_iceberg==0,name],range=[minval,maxval],\n",
    "             bins=nbins,color='r',alpha=0.5,label='Iceberg')\n",
    "    plt.legend()\n",
    "    plt.xlim([minval,maxval])\n",
    "    plt.xlabel(name)\n",
    "    plt.ylabel('Number')\n",
    "    plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAELCAYAAADDZxFQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBeucsWXLAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f6009725710>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nqrj/DzUwKvr/mDYdAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626b4f71d0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nCCSp4gwCSao4g0CSKs4gkKSK+/9qj4ya1FJz8QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626b366d68>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAEKCAYAAAAIO8L1AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626b20deb8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSTVnIpCkmvv/Y1U02WIACX4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626b34eda0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nFWcikKSK+/9lv/g0MSy1RAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626b3c1908>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYgAAAELCAYAAADDZxFQAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSaX+PwnjpHUcVnNsAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f626b329ac8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for col in ['inc_angle','min1','max1','std1','med1','mean1','mid50_1']:\n",
    "    plot_var(col)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Xtrain: (1604, 75, 75, 2)\n",
      "Ytrain: (1604,)\n",
      "Xtest: (8424, 75, 75, 2)\n"
     ]
    }
   ],
   "source": [
    "# Train data\n",
    "x_band1 = np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train[\"band_1\"]])\n",
    "x_band2 = np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in train[\"band_2\"]])\n",
    "X_train = np.concatenate([x_band1[:, :, :, np.newaxis], x_band2[:, :, :, np.newaxis]], axis=-1)\n",
    "y_train = np.array(train[\"is_iceberg\"])\n",
    "print(\"Xtrain:\", X_train.shape)\n",
    "print(\"Ytrain:\", y_train.shape)\n",
    "# Test data\n",
    "x_band1 = np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in test[\"band_1\"]])\n",
    "x_band2 = np.array([np.array(band).astype(np.float32).reshape(75, 75) for band in test[\"band_2\"]])\n",
    "X_test = np.concatenate([x_band1[:, :, :, np.newaxis], x_band2[:, :, :, np.newaxis]], axis=-1)\n",
    "print(\"Xtest:\", X_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Convolution2D, GlobalAveragePooling2D, Dense, Dropout"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 73, 73, 32)        608       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 71, 71, 64)        18496     \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d_1 ( (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 1)                 65        \n",
      "=================================================================\n",
      "Total params: 19,169\n",
      "Trainable params: 19,169\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "model = Sequential()\n",
    "model.add(Convolution2D(32, 3, activation=\"relu\", input_shape=(75, 75, 2)))\n",
    "model.add(Convolution2D(64, 3, activation=\"relu\", input_shape=(75, 75, 2)))\n",
    "model.add(GlobalAveragePooling2D())\n",
    "model.add(Dropout(0.3))\n",
    "model.add(Dense(1, activation=\"sigmoid\"))\n",
    "model.compile(\"adam\", \"binary_crossentropy\", metrics=[\"accuracy\"])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"470pt\" viewBox=\"0.00 0.00 504.00 470.00\" width=\"504pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(1 1) rotate(0) translate(4 466)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-466 500,-466 500,4 -4,4\" stroke=\"none\"/>\n",
       "<!-- 140060682410528 -->\n",
       "<g class=\"node\" id=\"node1\"><title>140060682410528</title>\n",
       "<polygon fill=\"none\" points=\"79,-415.5 79,-461.5 417,-461.5 417,-415.5 79,-415.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"165\" y=\"-434.8\">conv2d_1_input: InputLayer</text>\n",
       "<polyline fill=\"none\" points=\"251,-415.5 251,-461.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"278.5\" y=\"-446.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"251,-438.5 306,-438.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"278.5\" y=\"-423.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"306,-415.5 306,-461.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"361.5\" y=\"-446.3\">(None, 75, 75, 2)</text>\n",
       "<polyline fill=\"none\" points=\"306,-438.5 417,-438.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"361.5\" y=\"-423.3\">(None, 75, 75, 2)</text>\n",
       "</g>\n",
       "<!-- 140060682565336 -->\n",
       "<g class=\"node\" id=\"node2\"><title>140060682565336</title>\n",
       "<polygon fill=\"none\" points=\"100,-332.5 100,-378.5 396,-378.5 396,-332.5 100,-332.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"161.5\" y=\"-351.8\">conv2d_1: Conv2D</text>\n",
       "<polyline fill=\"none\" points=\"223,-332.5 223,-378.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"250.5\" y=\"-363.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"223,-355.5 278,-355.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"250.5\" y=\"-340.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"278,-332.5 278,-378.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"337\" y=\"-363.3\">(None, 75, 75, 2)</text>\n",
       "<polyline fill=\"none\" points=\"278,-355.5 396,-355.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"337\" y=\"-340.3\">(None, 73, 73, 32)</text>\n",
       "</g>\n",
       "<!-- 140060682410528&#45;&gt;140060682565336 -->\n",
       "<g class=\"edge\" id=\"edge1\"><title>140060682410528-&gt;140060682565336</title>\n",
       "<path d=\"M248,-415.366C248,-407.152 248,-397.658 248,-388.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"251.5,-388.607 248,-378.607 244.5,-388.607 251.5,-388.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140060682408680 -->\n",
       "<g class=\"node\" id=\"node3\"><title>140060682408680</title>\n",
       "<polygon fill=\"none\" points=\"100,-249.5 100,-295.5 396,-295.5 396,-249.5 100,-249.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"161.5\" y=\"-268.8\">conv2d_2: Conv2D</text>\n",
       "<polyline fill=\"none\" points=\"223,-249.5 223,-295.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"250.5\" y=\"-280.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"223,-272.5 278,-272.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"250.5\" y=\"-257.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"278,-249.5 278,-295.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"337\" y=\"-280.3\">(None, 73, 73, 32)</text>\n",
       "<polyline fill=\"none\" points=\"278,-272.5 396,-272.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"337\" y=\"-257.3\">(None, 71, 71, 64)</text>\n",
       "</g>\n",
       "<!-- 140060682565336&#45;&gt;140060682408680 -->\n",
       "<g class=\"edge\" id=\"edge2\"><title>140060682565336-&gt;140060682408680</title>\n",
       "<path d=\"M248,-332.366C248,-324.152 248,-314.658 248,-305.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"251.5,-305.607 248,-295.607 244.5,-305.607 251.5,-305.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140060682541712 -->\n",
       "<g class=\"node\" id=\"node4\"><title>140060682541712</title>\n",
       "<polygon fill=\"none\" points=\"-2.84217e-14,-166.5 -2.84217e-14,-212.5 496,-212.5 496,-166.5 -2.84217e-14,-166.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"161.5\" y=\"-185.8\">global_average_pooling2d_1: GlobalAveragePooling2D</text>\n",
       "<polyline fill=\"none\" points=\"323,-166.5 323,-212.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"350.5\" y=\"-197.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"323,-189.5 378,-189.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"350.5\" y=\"-174.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"378,-166.5 378,-212.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"437\" y=\"-197.3\">(None, 71, 71, 64)</text>\n",
       "<polyline fill=\"none\" points=\"378,-189.5 496,-189.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"437\" y=\"-174.3\">(None, 64)</text>\n",
       "</g>\n",
       "<!-- 140060682408680&#45;&gt;140060682541712 -->\n",
       "<g class=\"edge\" id=\"edge3\"><title>140060682408680-&gt;140060682541712</title>\n",
       "<path d=\"M248,-249.366C248,-241.152 248,-231.658 248,-222.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"251.5,-222.607 248,-212.607 244.5,-222.607 251.5,-222.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140060683147528 -->\n",
       "<g class=\"node\" id=\"node5\"><title>140060683147528</title>\n",
       "<polygon fill=\"none\" points=\"120,-83.5 120,-129.5 376,-129.5 376,-83.5 120,-83.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"182.5\" y=\"-102.8\">dropout_1: Dropout</text>\n",
       "<polyline fill=\"none\" points=\"245,-83.5 245,-129.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"272.5\" y=\"-114.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"245,-106.5 300,-106.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"272.5\" y=\"-91.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"300,-83.5 300,-129.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"338\" y=\"-114.3\">(None, 64)</text>\n",
       "<polyline fill=\"none\" points=\"300,-106.5 376,-106.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"338\" y=\"-91.3\">(None, 64)</text>\n",
       "</g>\n",
       "<!-- 140060682541712&#45;&gt;140060683147528 -->\n",
       "<g class=\"edge\" id=\"edge4\"><title>140060682541712-&gt;140060683147528</title>\n",
       "<path d=\"M248,-166.366C248,-158.152 248,-148.658 248,-139.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"251.5,-139.607 248,-129.607 244.5,-139.607 251.5,-139.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140060678580040 -->\n",
       "<g class=\"node\" id=\"node6\"><title>140060678580040</title>\n",
       "<polygon fill=\"none\" points=\"131.5,-0.5 131.5,-46.5 364.5,-46.5 364.5,-0.5 131.5,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"182.5\" y=\"-19.8\">dense_1: Dense</text>\n",
       "<polyline fill=\"none\" points=\"233.5,-0.5 233.5,-46.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"261\" y=\"-31.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"233.5,-23.5 288.5,-23.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"261\" y=\"-8.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"288.5,-0.5 288.5,-46.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"326.5\" y=\"-31.3\">(None, 64)</text>\n",
       "<polyline fill=\"none\" points=\"288.5,-23.5 364.5,-23.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"326.5\" y=\"-8.3\">(None, 1)</text>\n",
       "</g>\n",
       "<!-- 140060683147528&#45;&gt;140060678580040 -->\n",
       "<g class=\"edge\" id=\"edge5\"><title>140060683147528-&gt;140060678580040</title>\n",
       "<path d=\"M248,-83.3664C248,-75.1516 248,-65.6579 248,-56.7252\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"251.5,-56.6068 248,-46.6068 244.5,-56.6069 251.5,-56.6068\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(model, show_shapes = True).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pydot\n",
    "import graphviz\n",
    "from keras.utils import plot_model\n",
    "plot_model(model, to_file='model.png', show_shapes = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1203 samples, validate on 401 samples\n",
      "Epoch 1/10\n",
      "1203/1203 [==============================] - 2s 2ms/step - loss: 0.9586 - acc: 0.5777 - val_loss: 0.6383 - val_acc: 0.5985\n",
      "Epoch 2/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.6266 - acc: 0.6276 - val_loss: 0.6199 - val_acc: 0.5786\n",
      "Epoch 3/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5995 - acc: 0.6642 - val_loss: 0.6279 - val_acc: 0.5536\n",
      "Epoch 4/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5700 - acc: 0.6883 - val_loss: 0.6276 - val_acc: 0.5511\n",
      "Epoch 5/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5640 - acc: 0.6933 - val_loss: 0.6125 - val_acc: 0.5636\n",
      "Epoch 6/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5621 - acc: 0.6916 - val_loss: 0.6060 - val_acc: 0.5611\n",
      "Epoch 7/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5568 - acc: 0.7041 - val_loss: 0.6111 - val_acc: 0.5636\n",
      "Epoch 8/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5494 - acc: 0.7032 - val_loss: 0.6095 - val_acc: 0.5786\n",
      "Epoch 9/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5455 - acc: 0.7082 - val_loss: 0.6122 - val_acc: 0.5661\n",
      "Epoch 10/10\n",
      "1203/1203 [==============================] - 1s 1ms/step - loss: 0.5452 - acc: 0.7124 - val_loss: 0.5962 - val_acc: 0.5761\n"
     ]
    }
   ],
   "source": [
    "xz = model.fit(X_train, y_train, epochs=10, validation_split=0.25)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'acc': [0.57772236106203478,\n",
       "  0.62759767233681307,\n",
       "  0.66417290127881845,\n",
       "  0.68827930125016923,\n",
       "  0.69326683346271911,\n",
       "  0.69160432282429585,\n",
       "  0.70407315075447041,\n",
       "  0.70324189536093873,\n",
       "  0.70822942722666193,\n",
       "  0.71238570258405343],\n",
       " 'loss': [0.95857326407682275,\n",
       "  0.62663332859080534,\n",
       "  0.59945414181262968,\n",
       "  0.56996848275041145,\n",
       "  0.5640038461954715,\n",
       "  0.56206448402190745,\n",
       "  0.5567633269830039,\n",
       "  0.54938285497457706,\n",
       "  0.54551006354397769,\n",
       "  0.54521799374895108],\n",
       " 'val_acc': [0.59850374168885911,\n",
       "  0.57855361700057983,\n",
       "  0.55361596047135064,\n",
       "  0.55112219488531577,\n",
       "  0.56359102281549034,\n",
       "  0.56109725737809535,\n",
       "  0.56359102296413033,\n",
       "  0.57855361648033976,\n",
       "  0.5660847885501652,\n",
       "  0.5760598507456649],\n",
       " 'val_loss': [0.6382845467760081,\n",
       "  0.61992688369275328,\n",
       "  0.62793819610020174,\n",
       "  0.627625585494196,\n",
       "  0.61252939463256306,\n",
       "  0.60598422449425871,\n",
       "  0.61113002888876899,\n",
       "  0.60954373554695873,\n",
       "  0.61223516945827039,\n",
       "  0.59617643314704039]}"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "xz.history"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "batch_normalization_2 (Batch (None, 75, 75, 2)         8         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 73, 73, 8)         152       \n",
      "_________________________________________________________________\n",
      "max_pooling2d_5 (MaxPooling2 (None, 36, 36, 8)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_8 (Conv2D)            (None, 34, 34, 16)        1168      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_6 (MaxPooling2 (None, 17, 17, 16)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_9 (Conv2D)            (None, 15, 15, 32)        4640      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 7, 7, 32)          0         \n",
      "_________________________________________________________________\n",
      "conv2d_10 (Conv2D)           (None, 5, 5, 64)          18496     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2 (None, 2, 2, 64)          0         \n",
      "_________________________________________________________________\n",
      "global_max_pooling2d_2 (Glob (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 64)                4160      \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 64)                0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 32)                2080      \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 1)                 33        \n",
      "=================================================================\n",
      "Total params: 30,737\n",
      "Trainable params: 30,733\n",
      "Non-trainable params: 4\n",
      "_________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Conv2D, BatchNormalization, Dropout, MaxPooling2D, GlobalMaxPooling2D, Dense\n",
    "improved_cnn = Sequential()\n",
    "improved_cnn.add(BatchNormalization(input_shape = (75, 75, 2)))\n",
    "for i in range(4):\n",
    "    improved_cnn.add(Conv2D(8*2**i, kernel_size = (3,3)))\n",
    "    improved_cnn.add(MaxPooling2D((2,2)))\n",
    "improved_cnn.add(GlobalMaxPooling2D())\n",
    "improved_cnn.add(Dropout(0.2))\n",
    "improved_cnn.add(Dense(64))\n",
    "improved_cnn.add(Dropout(0.2))\n",
    "improved_cnn.add(Dense(32))\n",
    "improved_cnn.add(Dense(1, activation = 'sigmoid'))\n",
    "improved_cnn.compile(optimizer='adam', loss = 'binary_crossentropy', metrics = ['accuracy'])\n",
    "improved_cnn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/svg+xml": [
       "<svg height=\"1300pt\" viewBox=\"0.00 0.00 451.00 1300.00\" width=\"451pt\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n",
       "<g class=\"graph\" id=\"graph0\" transform=\"scale(1 1) rotate(0) translate(4 1296)\">\n",
       "<title>G</title>\n",
       "<polygon fill=\"white\" points=\"-4,4 -4,-1296 447,-1296 447,4 -4,4\" stroke=\"none\"/>\n",
       "<!-- 140059570540896 -->\n",
       "<g class=\"node\" id=\"node1\"><title>140059570540896</title>\n",
       "<polygon fill=\"none\" points=\"16.5,-1245.5 16.5,-1291.5 426.5,-1291.5 426.5,-1245.5 16.5,-1245.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"138.5\" y=\"-1264.8\">batch_normalization_2_input: InputLayer</text>\n",
       "<polyline fill=\"none\" points=\"260.5,-1245.5 260.5,-1291.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"288\" y=\"-1276.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"260.5,-1268.5 315.5,-1268.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"288\" y=\"-1253.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"315.5,-1245.5 315.5,-1291.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"371\" y=\"-1276.3\">(None, 75, 75, 2)</text>\n",
       "<polyline fill=\"none\" points=\"315.5,-1268.5 426.5,-1268.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"371\" y=\"-1253.3\">(None, 75, 75, 2)</text>\n",
       "</g>\n",
       "<!-- 140059570542576 -->\n",
       "<g class=\"node\" id=\"node2\"><title>140059570542576</title>\n",
       "<polygon fill=\"none\" points=\"8.5,-1162.5 8.5,-1208.5 434.5,-1208.5 434.5,-1162.5 8.5,-1162.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"138.5\" y=\"-1181.8\">batch_normalization_2: BatchNormalization</text>\n",
       "<polyline fill=\"none\" points=\"268.5,-1162.5 268.5,-1208.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"296\" y=\"-1193.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"268.5,-1185.5 323.5,-1185.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"296\" y=\"-1170.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"323.5,-1162.5 323.5,-1208.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"379\" y=\"-1193.3\">(None, 75, 75, 2)</text>\n",
       "<polyline fill=\"none\" points=\"323.5,-1185.5 434.5,-1185.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"379\" y=\"-1170.3\">(None, 75, 75, 2)</text>\n",
       "</g>\n",
       "<!-- 140059570540896&#45;&gt;140059570542576 -->\n",
       "<g class=\"edge\" id=\"edge1\"><title>140059570540896-&gt;140059570542576</title>\n",
       "<path d=\"M221.5,-1245.37C221.5,-1237.15 221.5,-1227.66 221.5,-1218.73\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-1218.61 221.5,-1208.61 218,-1218.61 225,-1218.61\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059570495160 -->\n",
       "<g class=\"node\" id=\"node3\"><title>140059570495160</title>\n",
       "<polygon fill=\"none\" points=\"77,-1079.5 77,-1125.5 366,-1125.5 366,-1079.5 77,-1079.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"138.5\" y=\"-1098.8\">conv2d_7: Conv2D</text>\n",
       "<polyline fill=\"none\" points=\"200,-1079.5 200,-1125.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"227.5\" y=\"-1110.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"200,-1102.5 255,-1102.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"227.5\" y=\"-1087.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"255,-1079.5 255,-1125.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"310.5\" y=\"-1110.3\">(None, 75, 75, 2)</text>\n",
       "<polyline fill=\"none\" points=\"255,-1102.5 366,-1102.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"310.5\" y=\"-1087.3\">(None, 73, 73, 8)</text>\n",
       "</g>\n",
       "<!-- 140059570542576&#45;&gt;140059570495160 -->\n",
       "<g class=\"edge\" id=\"edge2\"><title>140059570542576-&gt;140059570495160</title>\n",
       "<path d=\"M221.5,-1162.37C221.5,-1154.15 221.5,-1144.66 221.5,-1135.73\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-1135.61 221.5,-1125.61 218,-1135.61 225,-1135.61\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059570542296 -->\n",
       "<g class=\"node\" id=\"node4\"><title>140059570542296</title>\n",
       "<polygon fill=\"none\" points=\"35,-996.5 35,-1042.5 408,-1042.5 408,-996.5 35,-996.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"138.5\" y=\"-1015.8\">max_pooling2d_5: MaxPooling2D</text>\n",
       "<polyline fill=\"none\" points=\"242,-996.5 242,-1042.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"269.5\" y=\"-1027.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"242,-1019.5 297,-1019.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"269.5\" y=\"-1004.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"297,-996.5 297,-1042.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-1027.3\">(None, 73, 73, 8)</text>\n",
       "<polyline fill=\"none\" points=\"297,-1019.5 408,-1019.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-1004.3\">(None, 36, 36, 8)</text>\n",
       "</g>\n",
       "<!-- 140059570495160&#45;&gt;140059570542296 -->\n",
       "<g class=\"edge\" id=\"edge3\"><title>140059570495160-&gt;140059570542296</title>\n",
       "<path d=\"M221.5,-1079.37C221.5,-1071.15 221.5,-1061.66 221.5,-1052.73\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-1052.61 221.5,-1042.61 218,-1052.61 225,-1052.61\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603368144 -->\n",
       "<g class=\"node\" id=\"node5\"><title>140059603368144</title>\n",
       "<polygon fill=\"none\" points=\"73.5,-913.5 73.5,-959.5 369.5,-959.5 369.5,-913.5 73.5,-913.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"135\" y=\"-932.8\">conv2d_8: Conv2D</text>\n",
       "<polyline fill=\"none\" points=\"196.5,-913.5 196.5,-959.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"224\" y=\"-944.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"196.5,-936.5 251.5,-936.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"224\" y=\"-921.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"251.5,-913.5 251.5,-959.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"310.5\" y=\"-944.3\">(None, 36, 36, 8)</text>\n",
       "<polyline fill=\"none\" points=\"251.5,-936.5 369.5,-936.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"310.5\" y=\"-921.3\">(None, 34, 34, 16)</text>\n",
       "</g>\n",
       "<!-- 140059570542296&#45;&gt;140059603368144 -->\n",
       "<g class=\"edge\" id=\"edge4\"><title>140059570542296-&gt;140059603368144</title>\n",
       "<path d=\"M221.5,-996.366C221.5,-988.152 221.5,-978.658 221.5,-969.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-969.607 221.5,-959.607 218,-969.607 225,-969.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603368368 -->\n",
       "<g class=\"node\" id=\"node6\"><title>140059603368368</title>\n",
       "<polygon fill=\"none\" points=\"31.5,-830.5 31.5,-876.5 411.5,-876.5 411.5,-830.5 31.5,-830.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"135\" y=\"-849.8\">max_pooling2d_6: MaxPooling2D</text>\n",
       "<polyline fill=\"none\" points=\"238.5,-830.5 238.5,-876.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"266\" y=\"-861.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"238.5,-853.5 293.5,-853.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"266\" y=\"-838.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"293.5,-830.5 293.5,-876.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-861.3\">(None, 34, 34, 16)</text>\n",
       "<polyline fill=\"none\" points=\"293.5,-853.5 411.5,-853.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-838.3\">(None, 17, 17, 16)</text>\n",
       "</g>\n",
       "<!-- 140059603368144&#45;&gt;140059603368368 -->\n",
       "<g class=\"edge\" id=\"edge5\"><title>140059603368144-&gt;140059603368368</title>\n",
       "<path d=\"M221.5,-913.366C221.5,-905.152 221.5,-895.658 221.5,-886.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-886.607 221.5,-876.607 218,-886.607 225,-886.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059571185424 -->\n",
       "<g class=\"node\" id=\"node7\"><title>140059571185424</title>\n",
       "<polygon fill=\"none\" points=\"73.5,-747.5 73.5,-793.5 369.5,-793.5 369.5,-747.5 73.5,-747.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"135\" y=\"-766.8\">conv2d_9: Conv2D</text>\n",
       "<polyline fill=\"none\" points=\"196.5,-747.5 196.5,-793.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"224\" y=\"-778.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"196.5,-770.5 251.5,-770.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"224\" y=\"-755.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"251.5,-747.5 251.5,-793.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"310.5\" y=\"-778.3\">(None, 17, 17, 16)</text>\n",
       "<polyline fill=\"none\" points=\"251.5,-770.5 369.5,-770.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"310.5\" y=\"-755.3\">(None, 15, 15, 32)</text>\n",
       "</g>\n",
       "<!-- 140059603368368&#45;&gt;140059571185424 -->\n",
       "<g class=\"edge\" id=\"edge6\"><title>140059603368368-&gt;140059571185424</title>\n",
       "<path d=\"M221.5,-830.366C221.5,-822.152 221.5,-812.658 221.5,-803.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-803.607 221.5,-793.607 218,-803.607 225,-803.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059571184472 -->\n",
       "<g class=\"node\" id=\"node8\"><title>140059571184472</title>\n",
       "<polygon fill=\"none\" points=\"31.5,-664.5 31.5,-710.5 411.5,-710.5 411.5,-664.5 31.5,-664.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"135\" y=\"-683.8\">max_pooling2d_7: MaxPooling2D</text>\n",
       "<polyline fill=\"none\" points=\"238.5,-664.5 238.5,-710.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"266\" y=\"-695.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"238.5,-687.5 293.5,-687.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"266\" y=\"-672.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"293.5,-664.5 293.5,-710.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-695.3\">(None, 15, 15, 32)</text>\n",
       "<polyline fill=\"none\" points=\"293.5,-687.5 411.5,-687.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-672.3\">(None, 7, 7, 32)</text>\n",
       "</g>\n",
       "<!-- 140059571185424&#45;&gt;140059571184472 -->\n",
       "<g class=\"edge\" id=\"edge7\"><title>140059571185424-&gt;140059571184472</title>\n",
       "<path d=\"M221.5,-747.366C221.5,-739.152 221.5,-729.658 221.5,-720.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-720.607 221.5,-710.607 218,-720.607 225,-720.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603309904 -->\n",
       "<g class=\"node\" id=\"node9\"><title>140059603309904</title>\n",
       "<polygon fill=\"none\" points=\"77,-581.5 77,-627.5 366,-627.5 366,-581.5 77,-581.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"142\" y=\"-600.8\">conv2d_10: Conv2D</text>\n",
       "<polyline fill=\"none\" points=\"207,-581.5 207,-627.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-612.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"207,-604.5 262,-604.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-589.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"262,-581.5 262,-627.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"314\" y=\"-612.3\">(None, 7, 7, 32)</text>\n",
       "<polyline fill=\"none\" points=\"262,-604.5 366,-604.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"314\" y=\"-589.3\">(None, 5, 5, 64)</text>\n",
       "</g>\n",
       "<!-- 140059571184472&#45;&gt;140059603309904 -->\n",
       "<g class=\"edge\" id=\"edge8\"><title>140059571184472-&gt;140059603309904</title>\n",
       "<path d=\"M221.5,-664.366C221.5,-656.152 221.5,-646.658 221.5,-637.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-637.607 221.5,-627.607 218,-637.607 225,-637.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603310912 -->\n",
       "<g class=\"node\" id=\"node10\"><title>140059603310912</title>\n",
       "<polygon fill=\"none\" points=\"38.5,-498.5 38.5,-544.5 404.5,-544.5 404.5,-498.5 38.5,-498.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"142\" y=\"-517.8\">max_pooling2d_8: MaxPooling2D</text>\n",
       "<polyline fill=\"none\" points=\"245.5,-498.5 245.5,-544.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"273\" y=\"-529.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"245.5,-521.5 300.5,-521.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"273\" y=\"-506.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"300.5,-498.5 300.5,-544.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-529.3\">(None, 5, 5, 64)</text>\n",
       "<polyline fill=\"none\" points=\"300.5,-521.5 404.5,-521.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"352.5\" y=\"-506.3\">(None, 2, 2, 64)</text>\n",
       "</g>\n",
       "<!-- 140059603309904&#45;&gt;140059603310912 -->\n",
       "<g class=\"edge\" id=\"edge9\"><title>140059603309904-&gt;140059603310912</title>\n",
       "<path d=\"M221.5,-581.366C221.5,-573.152 221.5,-563.658 221.5,-554.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-554.607 221.5,-544.607 218,-554.607 225,-554.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059570492584 -->\n",
       "<g class=\"node\" id=\"node11\"><title>140059570492584</title>\n",
       "<polygon fill=\"none\" points=\"0,-415.5 0,-461.5 443,-461.5 443,-415.5 0,-415.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"142\" y=\"-434.8\">global_max_pooling2d_2: GlobalMaxPooling2D</text>\n",
       "<polyline fill=\"none\" points=\"284,-415.5 284,-461.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"311.5\" y=\"-446.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"284,-438.5 339,-438.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"311.5\" y=\"-423.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"339,-415.5 339,-461.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"391\" y=\"-446.3\">(None, 2, 2, 64)</text>\n",
       "<polyline fill=\"none\" points=\"339,-438.5 443,-438.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"391\" y=\"-423.3\">(None, 64)</text>\n",
       "</g>\n",
       "<!-- 140059603310912&#45;&gt;140059570492584 -->\n",
       "<g class=\"edge\" id=\"edge10\"><title>140059603310912-&gt;140059570492584</title>\n",
       "<path d=\"M221.5,-498.366C221.5,-490.152 221.5,-480.658 221.5,-471.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-471.607 221.5,-461.607 218,-471.607 225,-471.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059602948168 -->\n",
       "<g class=\"node\" id=\"node12\"><title>140059602948168</title>\n",
       "<polygon fill=\"none\" points=\"93.5,-332.5 93.5,-378.5 349.5,-378.5 349.5,-332.5 93.5,-332.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"156\" y=\"-351.8\">dropout_4: Dropout</text>\n",
       "<polyline fill=\"none\" points=\"218.5,-332.5 218.5,-378.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"246\" y=\"-363.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"218.5,-355.5 273.5,-355.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"246\" y=\"-340.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"273.5,-332.5 273.5,-378.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"311.5\" y=\"-363.3\">(None, 64)</text>\n",
       "<polyline fill=\"none\" points=\"273.5,-355.5 349.5,-355.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"311.5\" y=\"-340.3\">(None, 64)</text>\n",
       "</g>\n",
       "<!-- 140059570492584&#45;&gt;140059602948168 -->\n",
       "<g class=\"edge\" id=\"edge11\"><title>140059570492584-&gt;140059602948168</title>\n",
       "<path d=\"M221.5,-415.366C221.5,-407.152 221.5,-397.658 221.5,-388.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-388.607 221.5,-378.607 218,-388.607 225,-388.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603339640 -->\n",
       "<g class=\"node\" id=\"node13\"><title>140059603339640</title>\n",
       "<polygon fill=\"none\" points=\"105,-249.5 105,-295.5 338,-295.5 338,-249.5 105,-249.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"156\" y=\"-268.8\">dense_5: Dense</text>\n",
       "<polyline fill=\"none\" points=\"207,-249.5 207,-295.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-280.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"207,-272.5 262,-272.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-257.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"262,-249.5 262,-295.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300\" y=\"-280.3\">(None, 64)</text>\n",
       "<polyline fill=\"none\" points=\"262,-272.5 338,-272.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300\" y=\"-257.3\">(None, 64)</text>\n",
       "</g>\n",
       "<!-- 140059602948168&#45;&gt;140059603339640 -->\n",
       "<g class=\"edge\" id=\"edge12\"><title>140059602948168-&gt;140059603339640</title>\n",
       "<path d=\"M221.5,-332.366C221.5,-324.152 221.5,-314.658 221.5,-305.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-305.607 221.5,-295.607 218,-305.607 225,-305.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603339976 -->\n",
       "<g class=\"node\" id=\"node14\"><title>140059603339976</title>\n",
       "<polygon fill=\"none\" points=\"93.5,-166.5 93.5,-212.5 349.5,-212.5 349.5,-166.5 93.5,-166.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"156\" y=\"-185.8\">dropout_5: Dropout</text>\n",
       "<polyline fill=\"none\" points=\"218.5,-166.5 218.5,-212.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"246\" y=\"-197.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"218.5,-189.5 273.5,-189.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"246\" y=\"-174.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"273.5,-166.5 273.5,-212.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"311.5\" y=\"-197.3\">(None, 64)</text>\n",
       "<polyline fill=\"none\" points=\"273.5,-189.5 349.5,-189.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"311.5\" y=\"-174.3\">(None, 64)</text>\n",
       "</g>\n",
       "<!-- 140059603339640&#45;&gt;140059603339976 -->\n",
       "<g class=\"edge\" id=\"edge13\"><title>140059603339640-&gt;140059603339976</title>\n",
       "<path d=\"M221.5,-249.366C221.5,-241.152 221.5,-231.658 221.5,-222.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-222.607 221.5,-212.607 218,-222.607 225,-222.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059602962136 -->\n",
       "<g class=\"node\" id=\"node15\"><title>140059602962136</title>\n",
       "<polygon fill=\"none\" points=\"105,-83.5 105,-129.5 338,-129.5 338,-83.5 105,-83.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"156\" y=\"-102.8\">dense_6: Dense</text>\n",
       "<polyline fill=\"none\" points=\"207,-83.5 207,-129.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-114.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"207,-106.5 262,-106.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-91.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"262,-83.5 262,-129.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300\" y=\"-114.3\">(None, 64)</text>\n",
       "<polyline fill=\"none\" points=\"262,-106.5 338,-106.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300\" y=\"-91.3\">(None, 32)</text>\n",
       "</g>\n",
       "<!-- 140059603339976&#45;&gt;140059602962136 -->\n",
       "<g class=\"edge\" id=\"edge14\"><title>140059603339976-&gt;140059602962136</title>\n",
       "<path d=\"M221.5,-166.366C221.5,-158.152 221.5,-148.658 221.5,-139.725\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-139.607 221.5,-129.607 218,-139.607 225,-139.607\" stroke=\"black\"/>\n",
       "</g>\n",
       "<!-- 140059603320280 -->\n",
       "<g class=\"node\" id=\"node16\"><title>140059603320280</title>\n",
       "<polygon fill=\"none\" points=\"105,-0.5 105,-46.5 338,-46.5 338,-0.5 105,-0.5\" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"156\" y=\"-19.8\">dense_7: Dense</text>\n",
       "<polyline fill=\"none\" points=\"207,-0.5 207,-46.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-31.3\">input:</text>\n",
       "<polyline fill=\"none\" points=\"207,-23.5 262,-23.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"234.5\" y=\"-8.3\">output:</text>\n",
       "<polyline fill=\"none\" points=\"262,-0.5 262,-46.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300\" y=\"-31.3\">(None, 32)</text>\n",
       "<polyline fill=\"none\" points=\"262,-23.5 338,-23.5 \" stroke=\"black\"/>\n",
       "<text font-family=\"Times,serif\" font-size=\"14.00\" text-anchor=\"middle\" x=\"300\" y=\"-8.3\">(None, 1)</text>\n",
       "</g>\n",
       "<!-- 140059602962136&#45;&gt;140059603320280 -->\n",
       "<g class=\"edge\" id=\"edge15\"><title>140059602962136-&gt;140059603320280</title>\n",
       "<path d=\"M221.5,-83.3664C221.5,-75.1516 221.5,-65.6579 221.5,-56.7252\" fill=\"none\" stroke=\"black\"/>\n",
       "<polygon fill=\"black\" points=\"225,-56.6068 221.5,-46.6068 218,-56.6069 225,-56.6068\" stroke=\"black\"/>\n",
       "</g>\n",
       "</g>\n",
       "</svg>"
      ],
      "text/plain": [
       "<IPython.core.display.SVG object>"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SVG(model_to_dot(improved_cnn, show_shapes = True).create(prog='dot', format='svg'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 1283 samples, validate on 321 samples\n",
      "Epoch 1/30\n",
      "1283/1283 [==============================] - 2s 1ms/step - loss: 0.6454 - acc: 0.6033 - val_loss: 0.5568 - val_acc: 0.6822\n",
      "Epoch 2/30\n",
      "1283/1283 [==============================] - 1s 476us/step - loss: 0.5460 - acc: 0.7023 - val_loss: 0.4957 - val_acc: 0.6822\n",
      "Epoch 3/30\n",
      "1283/1283 [==============================] - 1s 475us/step - loss: 0.4333 - acc: 0.7825 - val_loss: 0.3754 - val_acc: 0.8162\n",
      "Epoch 4/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.3704 - acc: 0.8309 - val_loss: 0.4253 - val_acc: 0.7539\n",
      "Epoch 5/30\n",
      "1283/1283 [==============================] - 1s 473us/step - loss: 0.3261 - acc: 0.8542 - val_loss: 0.3295 - val_acc: 0.8131\n",
      "Epoch 6/30\n",
      "1283/1283 [==============================] - 1s 473us/step - loss: 0.3197 - acc: 0.8613 - val_loss: 0.3306 - val_acc: 0.8349\n",
      "Epoch 7/30\n",
      "1283/1283 [==============================] - 1s 470us/step - loss: 0.2424 - acc: 0.9002 - val_loss: 0.2968 - val_acc: 0.8505\n",
      "Epoch 8/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.2482 - acc: 0.9010 - val_loss: 0.2715 - val_acc: 0.8816\n",
      "Epoch 9/30\n",
      "1283/1283 [==============================] - 1s 475us/step - loss: 0.2166 - acc: 0.9182 - val_loss: 0.3742 - val_acc: 0.8287\n",
      "Epoch 10/30\n",
      "1283/1283 [==============================] - 1s 474us/step - loss: 0.1851 - acc: 0.9291 - val_loss: 0.4785 - val_acc: 0.8037\n",
      "Epoch 11/30\n",
      "1283/1283 [==============================] - 1s 474us/step - loss: 0.1349 - acc: 0.9525 - val_loss: 0.2915 - val_acc: 0.8692\n",
      "Epoch 12/30\n",
      "1283/1283 [==============================] - 1s 474us/step - loss: 0.0950 - acc: 0.9680 - val_loss: 0.3197 - val_acc: 0.8816\n",
      "Epoch 13/30\n",
      "1283/1283 [==============================] - 1s 475us/step - loss: 0.1106 - acc: 0.9618 - val_loss: 0.5120 - val_acc: 0.7975\n",
      "Epoch 14/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.0930 - acc: 0.9688 - val_loss: 0.4493 - val_acc: 0.8037\n",
      "Epoch 15/30\n",
      "1283/1283 [==============================] - 1s 471us/step - loss: 0.0465 - acc: 0.9836 - val_loss: 0.3477 - val_acc: 0.8816\n",
      "Epoch 16/30\n",
      "1283/1283 [==============================] - 1s 476us/step - loss: 0.1104 - acc: 0.9602 - val_loss: 1.5519 - val_acc: 0.6729\n",
      "Epoch 17/30\n",
      "1283/1283 [==============================] - 1s 473us/step - loss: 0.1681 - acc: 0.9470 - val_loss: 0.4843 - val_acc: 0.8224\n",
      "Epoch 18/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.0488 - acc: 0.9860 - val_loss: 0.4216 - val_acc: 0.8411\n",
      "Epoch 19/30\n",
      "1283/1283 [==============================] - 1s 476us/step - loss: 0.0225 - acc: 0.9953 - val_loss: 0.4434 - val_acc: 0.8442\n",
      "Epoch 20/30\n",
      "1283/1283 [==============================] - 1s 471us/step - loss: 0.0138 - acc: 0.9977 - val_loss: 0.5242 - val_acc: 0.8318\n",
      "Epoch 21/30\n",
      "1283/1283 [==============================] - 1s 474us/step - loss: 0.0129 - acc: 0.9977 - val_loss: 0.6010 - val_acc: 0.8131\n",
      "Epoch 22/30\n",
      "1283/1283 [==============================] - 1s 473us/step - loss: 0.0339 - acc: 0.9883 - val_loss: 0.6942 - val_acc: 0.7975\n",
      "Epoch 23/30\n",
      "1283/1283 [==============================] - 1s 473us/step - loss: 0.0184 - acc: 0.9961 - val_loss: 0.6380 - val_acc: 0.8069\n",
      "Epoch 24/30\n",
      "1283/1283 [==============================] - 1s 473us/step - loss: 0.0197 - acc: 0.9969 - val_loss: 0.6141 - val_acc: 0.8162\n",
      "Epoch 25/30\n",
      "1283/1283 [==============================] - 1s 471us/step - loss: 0.0499 - acc: 0.9821 - val_loss: 1.4188 - val_acc: 0.7040\n",
      "Epoch 26/30\n",
      "1283/1283 [==============================] - 1s 471us/step - loss: 0.0784 - acc: 0.9782 - val_loss: 0.5905 - val_acc: 0.8224\n",
      "Epoch 27/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.0399 - acc: 0.9883 - val_loss: 0.6638 - val_acc: 0.8193\n",
      "Epoch 28/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.0243 - acc: 0.9906 - val_loss: 0.5967 - val_acc: 0.8318\n",
      "Epoch 29/30\n",
      "1283/1283 [==============================] - 1s 472us/step - loss: 0.0051 - acc: 0.9992 - val_loss: 0.5681 - val_acc: 0.8411\n",
      "Epoch 30/30\n",
      "1283/1283 [==============================] - 1s 471us/step - loss: 0.0048 - acc: 0.9992 - val_loss: 0.5279 - val_acc: 0.8629\n"
     ]
    }
   ],
   "source": [
    "history1 = improved_cnn.fit(X_train, y_train, validation_split=0.2, epochs = 30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_acc', 'val_loss', 'acc', 'loss'])\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEWCAYAAACXGLsWAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f622af16ef0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEWCAYAAACJ0YulAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzt3Xd8lGXW8PHfSQjpCS2E3kEpCkiwomIH7A1FsSvqrq7u\nuq66665ln30fny2uaxeVtYso4rIKgl1UEAKCAtIFE0InlSSknfePayaEkDKTzGQyyfl+yCeZe+5y\n3UwyZ652LlFVjDHGGICIUBfAGGNM82FBwRhjTCULCsYYYypZUDDGGFPJgoIxxphKFhSMMcZUsqBg\njI9E5CUR+R8f990sIqc39jzGNDULCsYYYypZUDDGGFPJgoJpUTzNNneLyPcisk9EXhSRVBGZKyL5\nIvKxiLSvsv95IrJKRHJE5HMRGVzluZEissxz3FtATLVrnSMiyz3HfiMiRzawzDeJyAYR2Ssis0Wk\nm2e7iMg/RWSniOSJyA8iMszz3AQRWe0p21YR+W2D/sOMqcaCgmmJLgbOAAYB5wJzgd8DKbjf+V8B\niMgg4E3gTs9zc4D/ikhbEWkLvAe8CnQA3vacF8+xI4FpwM1AR+A5YLaIRPtTUBE5FfhfYCLQFdgC\nTPc8fSZwkuc+kj377PE89yJws6omAsOAT/25rjG1saBgWqInVHWHqm4FFgDfqup3qloMzAJGeva7\nDPhAVT9S1VLg70AscDxwLBAFPKaqpar6DrCkyjWmAM+p6reqWq6qLwP7Pcf540pgmqouU9X9wH3A\ncSLSBygFEoHDAVHVH1V1m+e4UmCIiCSparaqLvPzusbUyIKCaYl2VPm5qIbHCZ6fu+E+mQOgqhVA\nBtDd89xWPThj5JYqP/cG7vI0HeWISA7Q03OcP6qXoQBXG+iuqp8CTwJPATtFZKqIJHl2vRiYAGwR\nkS9E5Dg/r2tMjSwomNYsC/fmDrg2fNwb+1ZgG9Dds82rV5WfM4C/qGq7Kl9xqvpmI8sQj2uO2gqg\nqo+r6ihgCK4Z6W7P9iWqej7QGdfMNcPP6xpTIwsKpjWbAZwtIqeJSBRwF64J6BtgIVAG/EpEokTk\nIuDoKsc+D9wiIsd4OoTjReRsEUn0swxvAteJyAhPf8T/wzV3bRaR0Z7zRwH7gGKgwtPncaWIJHua\nvfKAikb8PxhTyYKCabVUdS0wGXgC2I3rlD5XVUtUtQS4CLgW2Ivrf3i3yrHpwE245p1sYINnX3/L\n8DHwR2AmrnbSH7jc83QSLvhk45qY9gB/8zx3FbBZRPKAW3B9E8Y0mtgiO8YYY7yspmCMMaaSBQVj\njDGVLCgYY4ypZEHBGGNMpTahLoC/OnXqpH369Al1MYwxJqwsXbp0t6qm1Ldf0IKCiEwDzgF2quqw\nWvYZCzyGSyewW1VPru+8ffr0IT09PZBFNcaYFk9EttS/V3Cbj14CxtX2pIi0A54GzlPVocClQSyL\nMcYYHwQtKKjql7hJP7W5AnhXVX/27L8zWGUxxhjjm1B2NA8C2nty2C8Vkatr21FEpohIuoik79q1\nqwmLaIwxrUsoO5rbAKOA03DpiheKyCJVXVd9R1WdCkwFSEtLO2QKdmlpKZmZmRQXFwe5yKEXExND\njx49iIqKCnVRjDEtUCiDQiawR1X3AftE5EtgOHBIUKj3RJmZJCYm0qdPHw5OatmyqCp79uwhMzOT\nvn37hro4xpgWKJTNR/8BxohIGxGJA44BfmzIiYqLi+nYsWOLDggAIkLHjh1bRY3IGBMawRyS+iYw\nFugkIpnAA7ihp6jqs6r6o4h8CHyPS/v7gqqubMT1Gl/oMNBa7tMYExpBCwqqOsmHff7GgVTAxrQ8\nGz+Fdr2hY/9Ql8QYn1iaiwDIycnh6aef9vu4CRMmkJOTE4QSmWZj5o3w1T9DXQpjfGZBIQBqCwpl\nZWV1HjdnzhzatWsXrGKZUKuogMK9sG93qEtijM/CLvdRc3TvvfeyceNGRowYQVRUFDExMbRv3541\na9awbt06LrjgAjIyMiguLuaOO+5gypQpwIGUHQUFBYwfP54xY8bwzTff0L17d/7zn/8QGxsb4jsz\njbI/D1AoqmsOpzHNS4sLCg/9dxWrs/ICes4h3ZJ44NyhtT7/yCOPsHLlSpYvX87nn3/O2WefzcqV\nKyuHjU6bNo0OHTpQVFTE6NGjufjii+nYseNB51i/fj1vvvkmzz//PBMnTmTmzJlMnjw5oPdhmlhR\ntvteuCe05TDGDy0uKDQHRx999EHzCB5//HFmzZoFQEZGBuvXrz8kKPTt25cRI0YAMGrUKDZv3txk\n5TVBUuzpL7KgYMJIiwsKdX2ibyrx8fGVP3/++ed8/PHHLFy4kLi4OMaOHVvjPIPo6OjKnyMjIykq\nKmqSspogKso58L2iHCIiQ1seY3xgHc0BkJiYSH5+fo3P5ebm0r59e+Li4lizZg2LFi1q4tKZkPE2\nH6EHAoQxzVyLqymEQseOHTnhhBMYNmwYsbGxpKamVj43btw4nn32WQYPHsxhhx3GscceG8KSmiZV\nXCUQFO6B+I6172tMM2FBIUDeeOONGrdHR0czd+7cGp/z9ht06tSJlSsPTOb+7W9/G/DymRCorClg\nI5BM2LDmI2OCpahaTcGYMGBBwZhgKcoGPLmqLCiYMGFBwZhgKc6Bdj3dzxYUTJiwoGBMsBTlQGI3\naBPr0l0YEwYsKBgTLEU5ENse4jpYUDBhw4KCMcFSlA2x7TxBwZqPTHiwoBAADU2dDfDYY49RWFgY\n4BKZZqHYW1PoaENSTdgIWlAQkWkislNE6lxNTURGi0iZiFwSrLIEmwUFc4jyUigpgJh2EGs1hSa3\nbw+smhXqUoSlYE5eewl4Enilth1EJBL4P2B+EMsRdFVTZ59xxhl07tyZGTNmsH//fi688EIeeugh\n9u3bx8SJE8nMzKS8vJw//vGP7Nixg6ysLE455RQ6derEZ599FupbMYHinaPgrSlYUGhay16GTx6C\nPifZTHI/BXM5zi9FpE89u90OzARGB+zCc++F7T8E7HQAdDkCxj9S69NVU2fPnz+fd955h8WLF6Oq\nnHfeeXz55Zfs2rWLbt268cEHHwAuJ1JycjKPPvoon332GZ06dQpsmU1oeVNcxLbzNB9ZUrwmlZfl\nvudnWVDwU8j6FESkO3Ah8EyoyhAM8+fPZ/78+YwcOZKjjjqKNWvWsH79eo444gg++ugj7rnnHhYs\nWEBycnKoi2qCyZviwjv6yJLiNa38bZ7vO0JbjjAUytxHjwH3qGqFiNS5o4hMAaYA9OrVq+6z1vGJ\nvimoKvfddx8333zzIc8tW7aMOXPmcP/993Paaafxpz/9KQQlNE3CGwBiPDUFsKR4TckbFAq2h7Yc\nYSiUo4/SgOkishm4BHhaRC6oaUdVnaqqaaqalpKS0pRl9EnV1NlnnXUW06ZNo6CgAICtW7eyc+dO\nsrKyiIuLY/Lkydx9990sW7bskGNNC1JZU/AMSQXrV2hK+Z5g4A0OxmchqymoauXSZCLyEvC+qr4X\nqvI0RtXU2ePHj+eKK67guOOOAyAhIYHXXnuNDRs2cPfddxMREUFUVBTPPONazaZMmcK4cePo1q2b\ndTS3JMVVOppLPaPLbFhq06ioqBIUrPnIX0ELCiLyJjAW6CQimcADQBSAqj4brOuGSvXU2XfcccdB\nj/v3789ZZ511yHG33347t99+e1DLZkLAW1OISYZSzyp6VlNoGoW7Qcvdz9Z85Ldgjj6a5Me+1war\nHMaERFEOtE2EyKiD+xRM8FVtMsq3oOAvm9FsTDB4U1wAtI2zpHhNyRsIOg6w5qMGaDFBQVVDXYQm\n0VruM+wV5xwICmBJ8ZqSt6bQbaRrPrK/Gb+0iKAQExPDnj17WvwbpqqyZ88eYmJiQl0UU5+ibDcc\n1cuS4jWd/O2AQJcjobzk4GVRTb1axBrNPXr0IDMzk127doW6KEEXExNDjx49Ql0MU5+iHEg57MBj\nS3XRdPKyID4Fkj1/J/nbDwwLNvVqEUEhKiqKvn371r+jMU2levNRbAfIzQxdeVqT/O2Q2MV9gWtO\nSh0S2jKFkRbRfGRMs6Lq6Whuf2Cb1RSaTv42SOwKCanucYF1NvvDgoIxgVZa5NqyD+pTqJIUzwRX\nTTUF4zMLCsYEWtVkeF6WFK9plJfCvl2uptA2HqKTbFiqnywoGBNoVdNme9kEtqZRsBPQA7WExC42\nq9lPFhSMCbRaawpYUAg278S1pG7ue0KqzWr2kwUFYwKtatpsL29NwZLiBZe3/6BqTcGCgl8sKBgT\naFXTZnvFWk2hSVQGha6e756g0MIntgaSBQVjAq1q2mwv61NoGvnbQCIhzrO8bUIXKN9/4DUx9bKg\nYEygFWWDRLgsqV6VSfEsKASVdzhqhOetrXJYqo1A8pUFBWMCrSjH9SdEVPvziusAhZaHJ6jytx0I\nBGBzFRrAgoIxgVY1bXZVlhQv+PK3H+hPANd8BDar2Q8WFIwJtOKcg/sTvCzVRfAdUlNIPbDd+CRo\nQUFEponIThFZWcvzV4rI9yLyg4h8IyLDg1UWY5qUt/moutgONiQ1mEqLXS2tak0hOhHaJlifgh+C\nWVN4CRhXx/M/ASer6hHAn4GpQSyLMU2nejI8L6spBJd35nLVoAA2q9lPQQsKqvolUOvHIlX9RlW9\nvW6LAFskwLQM1dNme1lSvODKqzZxzSvBJrD5o7n0KdwAzK3tSRGZIiLpIpLeGhbSMWGsosK98ddY\nU7CkeEFVfeKaV6KluvBHyIOCiJyCCwr31LaPqk5V1TRVTUtJSWm6whnjr/15gNbcp2AT2ILL+8Zf\nvaaQ2NWNPrJZzT4JaVAQkSOBF4DzVdX+Ukz4qykZnpclxQuu/G0QGX3o/31CKpQWegK2qU/IgoKI\n9ALeBa5S1XWhKocxAVVT2mwvqykEl3c2s8jB273NSTYCySdBW6NZRN4ExgKdRCQTeACIAlDVZ4E/\nAR2Bp8W9iGWqmhas8hjTJLw1hdqGpIINSw0W7zKc1VWdq5AyqGnLFIaCFhRUdVI9z98I3Bis6xsT\nEkU1JMPzsppCcOVvhy7DDt1us5r9EvKOZmNalJrSZntZUrzgqp7iwqsy/5GNQPKFBQVjAqmmtNlV\nWVK84NifDyX5h448AjerOSrOgoKPLCgYE0hFOW4ETFRszc9bUrzgyK9lNjO4jueEVJvV7CMLCsYE\nUm0pLrws1UVwVF+Gs7rErlZT8JEFBWMCqbYUF16xVlMIirpqCmCzmv1gQcGYQKotxYVXXEcbkhoM\nvtQUbPSRTywoGBNItaXN9vImxSsva7oytQb5293yp9GJNT+fkAolBa5D2tTJgoIxgVRvn4InKZ4t\nJB9Y1RfXqc7WavaZBQVjAqm+PoXKCWzWhBRQ3hQXtbG1mn1mQcGYQCkvdU0U9dYUsM7mQMvLqr2T\nGWxWsx8sKBgTKN4UF3X1KcRaUAg4VT9qCjYCqT4WFIwJlLrSZnt5m49sBFLgFGVD+f66awoxydAm\nxpqPfGBBwZhAqStttpclxQu82hbXqapyVrM1H9XHgoIxgVJX2mwvS4oXeLUtw1mdzWr2iQUFYwKl\nrrTZVcV1sNFHgeR9o0+qLyjYrGZfWFAwJlB8aT4CCwqB5q0pJNTRfAQ2q9lHFhSMCRRfmo/AkuIF\nWv52VzuLiql7v4RUt05zyb6mKVeYClpQEJFpIrJTRFbW8ryIyOMiskFEvheRo4JVFmOaRFGOS7UQ\nWc+ChpYUL7BqW4azOhuW6pNg1hReAsbV8fx4YKDnawrwTBDLYkzw1ZfiwsuS4gVWfSkuvBJtApsv\nghYUVPVLoK7f/POBV9RZBLQTER/CvTHNVHEOxCbXv58lxQus2pbhrC7BUl34IpR9Ct2BjCqPMz3b\nDiEiU0QkXUTSd+3a1SSFM8ZvPtcULClewFRU1D+b2cuS4vkkLDqaVXWqqqapalpKSkqoi2NMzepL\nm+1lSfECp3A3aLlvNYXY9m6pVKsp1CmUQWEr0LPK4x6ebcaEJ79qClhncyD4OnENbFazj0IZFGYD\nV3tGIR0L5KqqhXATnlTrT5vtZUnxAqe+ZTirS+xio4/qUc/YuYYTkTeBsUAnEckEHgCiAFT1WWAO\nMAHYABQC1wWrLMYEXWkhlJf4PvoILCgEQn3LcFaXmAq71gWvPC1A0IKCqk6q53kFfhms6xvTpHxJ\nm+1lmVIDJ28bIJDQ2bf9E7rAT18GtUjhLiw6mo1p9nxNcQGWFC+Q8rdBfApERvm2f2IXKM6F0qLg\nliuMWVAwJhB8WUuhKst/FBi+Dkf1slnN9bKgYEwg+NN8BBYUAsXXFBdeNqu5XhYUjAkEv2sKlhQv\nIPK3158yuyqb1VwvCwrGBII/fQpgSfECobwU9u1qWE3BZjXXyoKCMYFQlA0SCdFJvu1vNYXGK9gJ\nqH99CrEdICIKCqxPoTYWFIwJhKIctzi8iG/7x3V0o2AsKV7D+TOb2Ssiws1qto7mWllQMCYQfE1x\n4WVJ8RrP34lrXjaruU4WFIwJBF9TXHjZrObG8zfFhZcFhTpZUDAmEBpUU8CGpTZG/jbXjxPXyb/j\nElKtT6EOFhSMCQRf02Z7WVK8xvNOXIvw820ssasL4mX7g1OuMGdBwZhA8LumYM1HjebvxDWvxFTP\n8VZbqIkFBWMaq6LCjSRqSJ+CJcVrOH9TXHh5A4nNaq6RBQVjGmt/HqD+1RQsKV7jNbSmkOCtKdis\n5ppYUDCmsbwpLvzpUwDLf9QYpUXu/71BNQWb1VwXCwrGNJa/KS684izVRYM1dDgquNFKEmkjkGrh\nU1AQkTtEJMmzdOaLIrJMRM704bhxIrJWRDaIyL01PN9LRD4Tke9E5HsRmdCQmzAmpPxNhucV19Fq\nCg1VGRQaUFOwWc118rWmcL2q5gFnAu2Bq4BH6jpARCKBp4DxwBBgkogMqbbb/cAMVR0JXA487UfZ\njWke/E2b7WVJ8RquISkuqkq0oFAbX4OCN6HLBOBVVV1VZVttjgY2qOomVS0BpgPnV9tHAW8GsWQg\ny8fyGNN8NKqmYEGhQRpTUwAXTCwo1MjXoLBURObjgsI8EUkEKuo5pjuQUeVxpmdbVQ8Ck0UkE5gD\n3F7TiURkioiki0j6rl27fCyyMU2kwX0KlhSvwfK3QZsY/wOxl81qrpWvQeEG4F5gtKoWAlHAdQG4\n/iTgJVXtgacWIiKHlElVp6pqmqqmpaSkBOCyxgRQUbZ7g4qK9e84S4rXcN45Cr5mpa0usaurpZWV\nBLZcLYCvQeE4YK2q5ojIZFxfQG49x2wFelZ53MOzraobgBkAqroQiAH8TGRiTIj5m+LCy2Y1N1xD\n5yh4eWc12wS2Q/gaFJ4BCkVkOHAXsBF4pZ5jlgADRaSviLTFdSTPrrbPz8BpACIyGBcUrH3IhBd/\nU1x4WVK8hsvf1vD+BDiwLKcFhUP4GhTKVFVxHcVPqupTQGJdB6hqGXAbMA/4ETfKaJWIPCwi53l2\nuwu4SURWAG8C13quY0z48DfFhZclxWu4/O2NrCnYWs21aePjfvkich9uKOqJnnb/qPoOUtU5uA7k\nqtv+VOXn1cAJvhfXmGaoKBva9fL/OGs+apj9+VBS0LiaQmVQsM7m6nytKVwG7MfNV9iO6x/4W9BK\nZUw4sT6FptWY2cxe8SkgEdZ8VAOfgoInELwOJIvIOUCxqtbXp2BM61Cc07A+hbZxbtSSZUr1T2Mn\nrgFEREJ8Z2s+qoGvaS4mAouBS4GJwLcickkwC2ZMWCgvdU0ZDelTAEt10RCBqCmAZ1az1RSq87VP\n4Q+4OQo7AUQkBfgYeCdYBTMmLDQ0xYWXJcXzX2VNIbVx50nsCnnVR8kbX/sUIrwBwWOPH8ca03I1\nNMWFl9UU/Je/HdomQnSdAyDrZ0nxauRrTeFDEZmHGzYKruN5Th37G9M6NDTFhVdsB8jJqH8/c0Be\nVuNGHnkldoV9u10TYGS9gylbDZ+CgqreLSIXc2D46FRVnRW8YhkTJgJSU7DmI780dBnO6hJTAYWC\nnZBcPS1b6+VrTQFVnQnMDGJZjAk/je5TqJIUL9LnP8fmrygH3poMHQfAGQ9DTFL9x/gqfxv0PKbx\n56mc1bzdgkIVdfYLiEi+iOTV8JUvInlNVUhjmq1G1xRaYFK8sv0uIPy8EJa9DM+eAD8tCMy5VQNY\nU7BlOWtSZ1BQ1URVTarhK1FVAxj6jQlT3jfzmOSGHd/SJrBVVMB7v4DNC+D8p+C6DyGiDbx8Dnx4\nn1tbuTGKsqF8PyR1a3xZLdVFjWwEkTGNUZQN0UkNb/qJa2H5jz55CFa+A6f9CYZfDr2OgVu+gtE3\nwaKn4bmTYOvShp+/sYvrVBXfGRD/ZjW3gtRsFhSMaYyGprjwim1BmVIXPw9fPwZp18OY3xzY3jYe\nzv47XPUelOyDF86AT//SsLUMAjGb2SuyjUt34UtNoaIc5vwO/j4QNnzc+Gs3YxYUjGmM4pyGD0eF\nltN8tOYDmPs7GDQOxv+t5sVv+p8Ct34DR06EL/8KL5wGO1b7d51A1hTAt1nNpUXw9jWw+DmXL+n1\nS+GbJ1psraHVBIWVW3OZNHUROYW20pIJoKJsCwoZS+CdG6DrCLhkWt1NabHt4MJn4bLX3XyDqSfD\n1/9yn8R9ke9Zxj0hUEGha93LchbuhVcvhB/fh3GPwO3L4PBzYP79MOsWKC0OTDmakVYTFPaXlbN0\nSza/fGMZpeX1LS9tjI+KGpgMzyvck+Lt2QhvXuY+cV8xwzUV+WLwOfCLRTDwTPjoT/DC6bDoWdi7\nqe7j8re7/++omMaXHeqe1ZyTAdPGuT6QS6bBsbdCdAJc+jKc8gf4fjq8NAHyWlZHdasJCqN6d+Av\nFw7j6w17+PP7flZZjalNUXbj+hQgfFNd7NsNr13smlEmvwsJfq6fnpACl70GFz7n1kj48B54fCQ8\nkQYf/h42fX5ov0NjF9epLrEL7Nvl5olUtX0lvHiGu97kd2HYRQeei4iAk3/nyr5zDUwdC5npgStT\niLWg2TL1uzStJ+t25PP8gp8YlJrI5GN7h7pIJpypNjxtdlXhmBSvpBDemOg6aa95Hzr2b9h5RNwo\npeGXu1rH+o9g/XxY8gIsegraJkC/sTDoLBhwRuOX4awusQtohQsMSZ5g89OXMP1Kd+3r50Lq0JqP\nHXwu3NgP3pwE/54A5/4LRkwKXNlCJKhBQUTGAf8CIoEXVPWRGvaZCDwIKLBCVa8IZpnuHT+Y9TsL\neHD2KvqlxHN8/07BvJxpyUoLobykcX0KEH6pLirKYeYNsHWZ+7Tcc3Rgztuxv/s69hY3SumnL2Hd\nPBck1rzv9pEIGB7At4iqs5qTusLKma6voEM/mDwTknvUfXzqUJjyOcy4Gt67BXashNMfCuvZ6UFr\nPhKRSOApYDwwBJgkIkOq7TMQuA84QVWHAncGqzxekRHC45NG0qdTPL94fRlb9uwL9iVNS9XYFBde\nsR3Cp/lIFebcDWvnwIS/ub6BYGgbD4eNh3Mfg1+vcqOWTnsA+p4c2Gt6m6Lyd8DCp+Cd66F7Glz/\nYf0BwSuuA1w1C46eAgufhDcuPTDTPQwFs0/haGCDqm5S1RJgOnB+tX1uAp5S1WyAaum5gyYpJooX\nrk5DFW58OZ384tKmuKxpaRqb4sIrXGoKqvDJw5D+IpxwBxx9U9NcV8R9Ij/xN3D1ey5YBIp3TYZP\nHoJ5v4fB57k3eH9f08goFyTPfdyl9Hj+VNj4me+jqpqRYAaF7kDVnMCZnm1VDQIGicjXIrLI09x0\nCBGZIiLpIpK+a9eugBSuT6d4nrnyKDbt3scd05dTXtEyxxybIGps2myvuI7uXNU7O5uTigr44C74\n6lE46ho47cFQlygw4ju77ztXu0/6l77UuJFNo66Ba/4L+wvg1Qvg0cEw9x43bDdM5jWEevRRG2Ag\nMBaYBDwvIof8hanqVFVNU9W0lBQ/RzjU4fgBnXjw3CF8umYnf523JmDnNa1EwGoKnlnNzTUpXnkp\nzJpyoIZw7r/cCJyWoE1bOPJyOPMvMP6vbu3mxup9HNyxwgWYHqMhfRq8eDo8PsLVtPydsNfEgtkb\nshXoWeVxD8+2qjKBb1W1FPhJRNbhgsSSIJbrIFcd14e1O/J57otNDOqcyMWjfGxHNCZQfQpVJ7DF\nN7OBD6VF8Pa1sO5D16Z/4m/qPSTsXPRc4M/ZNg6GXui+inPd5LeV78BXj8GCf0DnITDsYjjiEmjf\nJ/DXb4RghvslwEAR6SsibYHLgdnV9nkPV0tARDrhmpPqmb0SeA+cO5Tj+nXkvnd/YNnP4dtBZJpY\noGsKza1foTgPXrvEjQA6+x8tMyA0hZhkGHml66u4ay1M+LtLovjpn+Ffw+HFs2Dt3GbTvBS0oKCq\nZcBtwDzgR2CGqq4SkYdF5DzPbvOAPSKyGvgMuFtVg/OXUVII306tseMnKjKCp688ii7JMUx5ZSlZ\nOY1M72tah+IckMjGrxXcHJPi7dsDr5wHGYvg4hdg9I2hLlHLkJDiOuhvmAd3/uCGr+Zvgzcvdxlk\nV892/TchFNSGQVWdo6qDVLW/qv7Fs+1Pqjrb87Oq6m9UdYiqHqGq04NWmJUzYe7drvOnhmnt7ePb\n8uI1aRSXlnPTK+kUljTjTj/TPBR5kuHVlPzNH80t/1FeFvx7POz8ES5/wzVxmMBr1wvG3Am3L4UL\nnnFzM2Zc5RYlWjkzZCOXWkhvkQ9GTnaLfmQsgWfHwMZPD9llYGoij08aweptedw78we0mVTnTDMV\niBQX0LyCwp6NMO0sFxgmz3QziU1wRUbBiCvgtiVw8YsuGLxzPTx1DKyY3uSj0lpPUBBxgWHKZ+6P\n8NWL4JM/H/Iffurhqfzm9EHMXpHF20szQ1RYExYCkeICDiTFC3VQ2L7SJYDbXwDX/hf6jAlteVqb\niEhXK/vFIjdyqU00zLoZnkyDZa+6UWBNUYwmuUpz0nkw3PSZ6/hZ8Hd4+VzIPXhQ1C9OGcCx/Trw\nwH9WsXFXQYgKapq9xqbNriquY2hnwWYsdhk/I9q42bzdRoauLK1dRIQbtXTzAtd8F5MEs2+Dx4+C\nH94J/uWHdJj7AAAgAElEQVSDfoXmqG2ca0q6cCpsW+Gak9bNr3w6MkJ47LKRxERFcPsb37G/LPxm\nJZom0Ni02VWFMine5q/hlQtcYLr+Q0g5LDTlMAeLiIDDz4YpX8AVb0NC5yaZy9I6g4LX8Mvg5i/c\nIuBvXArz/1hZReuSHMPfLhnO6m15PDLXJraFrZwM+PhBl+Y50ALVpwChS3Wx6Qt4/RKX5+e6udDe\nMgc3OyIw6Ey48WMYdV3QL9e6gwJAp4HuPzvtevjmcTfqIudnAE4fksq1x/fh319v5pMf/Vjc2zQf\nn/0/+Oqf8OyJ8PO3gTtvRYWblBSo5qNQJMXb+KlLf92uN1z7fmBTUpvAEwnMjOt6WFAAiIqFc/4J\nl/zbLZpRZXTSveMPZ3DXJO5+53t25LW8pfdatIJdbhbpwLNcOoOXJsDCpwMzSWh/LqABbD5q4prC\n+o/hjcuhQ38XEBI6N921TbNmQaGqYRfBLV9CUnd4+zrI20ZMVCRPTBpJUUk5d1rivPCy7CW33sGZ\nf3btsoPGwbz7XO774tzGnTtQKS68mjIp3rr5MH0SpAxyyduaW2oNE1IWFKrr0A8mvgpl+2H27aDK\ngM4JPHTeUBZu2sOzX2wMdQnDz661Td80Ul4KS6ZBv1Ncx2lsO7cgzJn/A2s+cEsobv+h4ecPVIoL\nL28K5yUvBDfdwdq5MP0KNwrv6tkQ3zF41zJhyYJCTToNgDMegg0fwbKXAbg0rQfnDu/Gox+tY+kW\ny4/ks63LXHv+jKub9rpr3of8LDjm5gPbROD42+HaD1yitxdOd+O/GyJQabO9jrjULWL/4T1uVbP9\n+YE5b1U//hfeugq6HAFX/+dAziVjqrCgUJvRN0Hfk2DeHyB7MyLCXy4cRrd2Mfzqze/ILbKFeeq1\nb48LBhVlsHlB4z6Z++vb51z2yYFnHvpc7+PcGPCex7jx3+/90uXG8oe3+ShQNYXoRJj0Fpz2J1g1\nC6ae4tJMBMqq91y2024j3EI1gSq3aXEsKNQmIgLOf9qtCfveL6CigqSYKB6/fCQ78or5/buWBqNO\n5WXwznVQsBMmvwNRcbDomaa59rbv4eeFLrDXNlojIcVlrTzpd7D8dVdr2L3B92t4m48C1acA7nfu\nxLvcp/jiXLd614oApANbOdOzzOQomPyuy9ppTC0sKNSlXU8Y9whs+RoWPQ3AyF7t+c2Zg/jgh228\ntSSjnhO0Yp/+GX76wqVc7n+qy+3yw9suSATb4udcEBo5ue79IiLh1D/Ale+4TJVTx8L6j3y7RqCb\nj6rqexLcssDNKp51M/z3Diht4Mi379+GmTe6WtHkmW52rDF1sKBQnxFXwGET3IpJO90ktltO6s+Y\nAZ148L+r2LAzCG2/4W71f+Drx9xEm6OuctuOucWNBFryYnCvvW+PeyMcfrnvb9gDT3dvwh36woxr\nYMeq+o8pynb5iqJiG1fe2iR2cR3BY34NS1+CF8+AvT4sNaLqajyLn4e3JrsV03qf4GprjU3xbVoF\nCwr1EYFzHoO28fDeLVBeSkSE8OjE4cS3bcNtb3xHTmFJqEvZfOxa65rbuqfB+P87sL3TQDdfIP3F\nhn/q9cWyl6F8v1tv1x/JPeCKGe6Nc/oV9Y+WCmSKi9pEtoHTH3R9DTk/w3Nj3Qpe1eXvgO9nuL6R\nfw6DJ0fBnN9C1nK3DsIVM9zvrzE+sKDgi8RUN7kt6ztY8CgAnZNiePSyEWzcVcA5T3zF95l15CQp\nL3XrtP60oIkKHCLFeTD9SvcJeuIrLstjVcfeCvs8E8qCobzM1UT6nuSGXPorqasbtpqX5fpD6poz\nEMgUF/U5bBzc/CV07AdvXQnz73eroX14Hzx9HPxjELx7E6z9AHqMcr+rty9zi7hM+JvL9WWMj4Ia\nFERknIisFZENInJvHftdLCIqImnBLE+jDL3ADRv88q8uOAAnD0rh7VuORxUueWYhry3acmjn85Zv\n3JDM938Nr1/qhmi2RKrw3q2uiePSlyC5+6H79Bvr1qZd9ExwxuKv/QDyMl1TVUP1HA1nPwqbPoeP\nH6h9v+Lcph3B0743XD/PffL/5gmXniJ9GiSkutW7pnwOd290wTjteujYv/GL/5hWKWhBQUQigaeA\n8cAQYJKIDKlhv0TgDiCAiWmCZMLfID4FZt1S2QQyomc73r99DMcP6Mj9763k128tZ9/+MpdiYdYt\nLpdSyT648Dk34uXNy12Stpbmq3+6uQFnPAx9T6x5HxFXW9ixEn76MvBl+HaqW81q0LjGneeoq1zz\n08Inax/9E8i02b5qE+067q+b6/ob7tnihpeOudN1SjdBXhzT8gWzpnA0sEFVN6lqCTAdOL+G/f4M\n/B/Q/BMLxbaH856EXWvgs/+p3Nw+vi3TrhnNb88cxPsrMnnhn3+g/IlRLvf5iXfBL791HZ9XvO2C\nyRsTXVNLS7HxUzfaaOhFcNwv6973iIkupUOgh6duXwlbvqp7GKo/zvp/0OdEmP2rmmt3TdGnUJve\nx0O/kyEqJjTXNy1aMINCd6DqR+JMz7ZKInIU0FNVP6jrRCIyRUTSRSR9165dgS+pPwaeDqOuhW+e\ndE1DHhERwm2H5fFd1//jjuJnWVLck09OmeUmI3nbdDsfDhNfht3r3ESiJl5mLyiyt8A7N0DK4XD+\nk/U3WUTFQNoNsO5Dt/RjoCx+DtrE1j8M1VeRUa4ZLCHVjeKpPpS2KfsUjGlCIetoFpEI4FHgrvr2\nVdWpqpqmqmkpKSnBL1x9zvwf10zx3q1u6cKibHj/N/D8aSSW7CJnwrP8o8tfueGDPP743sqDF+np\nf4prs974Ccy9O7h5boKttMgtNF5R5jpofR3hMvpGt8LXt88GphyFe90w1CMnBjZ1Q3wnuPx1d/63\nroIyzyizshIo3df0zUfGNIFgBoWtQM8qj3t4tnklAsOAz0VkM3AsMLtZdzZ7RSfChc+6T8nTr4An\n0txY8mNvhduW0O7oSbwx5TimnNSPVxdtYeKzC8nMrpJGYdQ1cMKdrqNw4VMhu41GUYUP7nIr1100\n1XVs+iox1a1F+93rB9JFNMayV6Cs6OA8R4HS9Ui44CnIWARzf+e2FQc4xYUxzUgwg8ISYKCI9BWR\ntsDlwGzvk6qaq6qdVLWPqvYBFgHnqWp6EMsUOL2Pd+3nP33h3hBv/gLG/W/ljNGoyAh+P2Ewz04e\nxaZd+zj78a/474qsA6OTTnsAhpzvhhfWNPa8uUt/0aWHOOl3cNh4/48/9lb3aXvZK40rR0W5G4ba\n50RIHdq4c9Vm2MUuiC/9twvkgU6bbUwzErSgoKplwG3APOBHYIaqrhKRh0XkvGBdt0md/hDc8DFc\n96HLPFmDccO68P6vxtCzQyy3v/kdFzz9DQs37nF5bi58Drof5dIQhNNQ1S0LYe49Ltnc2FpHGtet\n63DoPQYWT21c38rauZD7s/+T1fx12p9gwBkw525YN9dts5qCaYEk3JK6paWlaXp6eFQmqiqvUGYu\ny+SfH61jW24xYw9L4Z5xhzM4sRieP83Nwr3xE5dvqTnL3epyBEUnwE2fNa5dfc0Hrvnt0pdg6IUN\nO8dL50D2ZvjVcjcDOJiKclySur2bAHWvV4/m39ppDICILFXVen9hbUZzE4mMECam9eSz347lvvGH\ns2xLNhMeX8BvPshi2zmvuE7b5j5UtWy/61guLYTL32h8R+ugcS699cKnG3b8jtUuJffoG4IfEMDd\n76Q3oW2C57HVFEzLY0GhicVERXLzyf1Z8LtTmXJSPz74YRsnv7Sd13r9Gd21tv70CqHi7VjeuhQu\neKZhaSSqi4iEY26FzMWQ2YDa3+KpLqXGUdc0viy+SjkMLv23m52d3KPprmtME7HmoxDbllvEYx+t\n5+2lGVzd9nMelKmUHXEZbQ4b5zpRK8pAPd8ryt1X1cddhkG/U10fRTAtecEFhRN/C6f9MXDn3Z8P\njw5x/ROX+JFBtSjbHTfsYjc/whhTJ1+bj5qgzm3q0jU5lv+75EhuPLEvf52XyjPrtnHrD2/BD2/5\nfpL2fVy+mxGTg7PmbtWO5VN+H9hzRyfCUVe7OQu5D9ecM6km373mmrGCMQzVmFbMagrNTPrmvXz6\n9ULy9xWyt7iC7KJy9haVk1OslBNJORGUEUE5LpXDLV038IuEL4jM+AYio12H7egboMfowCREy8uC\n504OTMdybbK3wOMj4IQ7XKro2pTthw0fw8p3XSd1t5Fw/dzAl8eYFshqCmEqrU8H0vqcfcj20vIK\ncgpLyS4sYe++ErL3lbBhZwGPfRLHx5FjeOXaRJJXveoSuH0/3Q2RTbvBZXaNTmhYYcr2u5m8pYVw\nzezgzeBt3xsOPwfS/w0n3X3wzOjyUjcXZOW7bj7H/lyI7eBySY35dXDKY0wrZjWFMPfx6h384o1l\n9OkYx2s3HEPn6FK37OWSF1020ugk9waadr1/ncOqMPt2+O5VmPgqDAny1JItC+Hf41wKkFHXwc/f\nuLWFV/8HCve4+xh8rku61+9kl5vIGOMzX2sKFhRagG827ObGV9LplBDN6zceQ88Oce5NPWOx6yBe\n/Z5bCrPTILe06OFnu5XR6uqcruxYvstN3Ao2VXj+FDcPIiLSrZkcFedmSw+7GPqfZllBjWkECwqt\nzHc/Z3Ptv5cQGxXJazcezYDOVdbj3bfbfepe8wFs+dqNXIpPcfMEDj/bDa+sutbwloXw8jnQ7xS4\n4q2my9P/4/tuBbH+p8Kwi1z5bBlJYwLCgkIrtGZ7HpNfWEyFKq9cfzTDuicfulNRNqz/2K1Stv5j\nKMl3n8j7n+pqEV2Hw6sXBrdj2RjT5CwotFKbd+/jyhe+Ja+olBevHc3RfetIJV22380IXjPH5RDK\nz3Lbo+Lhpk8CM0HNGNMsWFBoxbJyipj84rdk5RTx7ORRjD2sc/0HqcK25bBuPvQ61nXmGmNaDMt9\n1Ip1axfLjJuPo1+nBG56JZ05P2yr/yARN+5/7D0WEIxpxWyeQgvVKSGaN6ccyw0vLeG2N5Zx/9lD\n6JcST25RKblFpeQUer6KSsgtLCWnqJScwhJyi8ronxLPxLSejD+iC3Ft7VfEmNbEmo9auMKSMm5+\ndSkL1u8+5LmE6DYkx0bRLs7zFduWhOg2fPvTHjbvKSQhug3nDu/GxLQejOjZDgnEDGljTEhYn4Kp\nVFJWQfqWvUS3iSA5ti3t4qJIjo0iKrLm1kNVZfFPe3krPYM5P2yjuLSCQakJTEzryYUju9MxIbqJ\n78AY01gWFExA5BeX8t8V25iRnsHyjByiIoXTB6cyMa0nJw1KITLCag/GhINmERREZBzwLyASeEFV\nH6n2/G+AG4EyYBdwvapuqeucFhRCZ92OfGYsyWDWd1vZs6+ELkkxXDyqO5eO6kmfTjbJzJjmLORB\nQUQigXXAGUAmsASYpKqrq+xzCvCtqhaKyK3AWFW9rK7zWlAIvZKyCj5ds5MZ6Rl8vnYnFQpH9+3A\nZdY5bUyz1RyCwnHAg6p6lufxfQCq+r+17D8SeFJVT6jrvBYUmpcdecXMXJbJ2+mZ/LR7n6dzuiuX\npvVkpHVOG9NsNIfU2d2BjCqPM4Fj6tj/BqDG5PgiMgWYAtCrV69Alc8EQGpSDL8YO4BbT+5P+pZs\n3lqSwXvfZfHm4gwGdE5gYloPLhzZg5RE65w2JhwEs6ZwCTBOVW/0PL4KOEZVb6th38nAbcDJqrq/\nrvNaTaH5K9hfxgffZzEjPZOlW7JpEyFcOLI7944/3EYuGRMizaGmsBXoWeVxD8+2g4jI6cAf8CEg\nmPCQEN2Gy0b34rLRvdiws4DXv93Cqwu3MH/1Du4ZdziXj+5JhI1aMqZZCmaaiyXAQBHpKyJtgcuB\n2VV38PQjPAecp6o7g1gWEyIDOifwwLlDmXvHiRzeJZHfz/qBi575hpVbc0NdtGbli3W7eOA/Kyku\nLQ91UUwrF7SgoKpluCahecCPwAxVXSUiD4uIdxmvvwEJwNsislxEZtdyOhPmBqYmMn3KsTw6cTiZ\n2YWc9+RXPDh7FfnFpaEuWkgVl5bz4OxVXDNtMS8v3MJfP1wb6iKZVs4mr5kml1tYyt/nr+W1b7eQ\nkhDNH88ZwjlHdm11I5VWbs3lzreWs2FnAded0If9ZRW88e3PvHHjMRw/oFOoi2damJAPSQ0WCwot\nx4qMHO5/byU/bM1lzIBOPHz+UPqlJIS6WEFXXqE8v2AT/5i/lg7xbfn7pcM5cWAKRSXlnP34AopL\ny/nw1yeRFGPrUJvAsdTZptkb3rMd7/3yBB4+fygrMnIY99gC/j5vLXktuEkpM7uQK55fxCNz13DG\nkFQ+vOMkThyYAkBs20gevWwEO/L389Ds1fWcyZjgsKBgQioyQrj6uD588tuTmXBEF578bAMnPPIp\nj85fS/a+klAXL6De+24r4x9bwKqsPP5x6XCeuuIo2se3PWifET3b8cux/Zm5LJMPV24PUUlNa2bN\nR6ZZWbk1lyc+Xc+8VTuIbxvJ5ON6c+OYfmE9+S23sJQ//mcls1dkkda7Pf+8bAQ9O8TVun9peQUX\nPv01WTnFzLvzpLC+d9N8WJ+CCWtrt+fz1GcbeP/7LNq2iWDS0b24+aT+dEmO8fkcxaXlrNuRz/bc\nYgZ3TaJH+9gm78xeuHEPd81Yzs78/fz6jEHccnJ/nzLLrt+Rz9lPfMVJAzvx/NVpra4T3gSeBQXT\nImzaVcDTn29k1ndbiRTh0rQe3HJy/0M+aecWlrJ6Wx6rsnJZnZXH6m15rN9ZQHnFgd/vTgltGd6j\nHSN6tmNEr3Yc2aMdybHB6cytqFCe+WIj/5i/lj4d43ns8hEc2aOdX+d4YcEm/ueDH/nrxUcycXTP\n+g8wpg4WFEyLkrG3kKc/38g7SzNQhQtGdqdn+zhWb8tlVVYemdlFlft2ToxmaLckhnZLZmi3JDon\nxbB6Wx7Lf85heUY2G3ftq9y3f0o8w3u2Y2TPdozo2Z4h3ZIavUZEblEpd81Ywcc/7uC84d3434uO\nID7a/+QBFRXKpOcXsXJrLh/eeVKdTU7G1MeCgmmRtuUW8dwXm3hz8c/sL6ugb6d4hnRLqgwCQ7om\n1dsGn1dcyvcZuSzPyGZ5Rg7LM3LYXeA6tQd0TuAPZw/mlMM6N6h8q7PyuPX1pWzNLuL+swdzzfF9\nGtX0k5ldyLjHFjCkWxLTbzrW0oOYBrOgYFq0/OJSRISEBnwCr05V2ZpTxLeb9vLEp+vZvKeQkwal\ncP/ZgxmUmujzeWYuzeQP7/1AcmwUT195FKN6d2h02QDeTs/g7ne+5w8TBnPTSf0Cck7jqCqb9xSS\nkhgdkN+l5syCgjENUFJWwSsLN/P4J+vZV1LOpKN78uvTB9WZ3XV/WTl/fn81ry36mWP7deCJSUcF\ndMSQqjLl1aV8sXYX/719DId18T1QmZpt3r2P2SuymL0iiw07C0iMacPVx/Xm2uP7ttjRXhYUjGmE\n7H0lPPbxOl779mfi2kZy+6kDuOb4PkS3iTxov6ycIm59fRkrMnK4+aR+3H3WYbSJDPz0n90F+znr\nn1/SJTmGWb84gbZtbIqRv7blFvH+im389/ssvs90CRmP7tOBccO6kL5lL3NXbqdtZAQT03py04n9\n6NWxZfXhWFAwJgA27MznLx/8yGdrd9G7Yxz3jR/MWUNTERG+Wr+bX03/jpKyCv5+6ZGMG9Y1qGWZ\nt2o7N7+6lNtOGcBvzzosqNdqKfbuK2HOD9uYvSKLJZv3ogrDuidx3vBunHNkN7q1i63cd9OuAqZ+\nuYmZyzIpr1DOObIbt5zcnyHdkkJ4B4FjQcGYAPpy3S7+54PVrNtRwDF9OzCyV3umfrmRAZ0TeGby\nKPo3Uc6m3769gneXZfKvy0dy5tDUQ2ouBopKypm/ejuzvtvKV+t3U1ah9E+J57zh3Tl3eNd682tt\nzy1m2tc/8fqiLewrKeeUw1K4dewARvdpH9bzRSwoGBNgZeUVTF+SwaMfrWPvvpJGDTdtqLziUs5/\n8mt+2r2PxOg2nDq4M+OGduHkw1KIa9uyO0rrUlGhfPvTXt5dlsncldsp2F9Gt+QYzh3RjfOGd2NI\n1yS/39BzC0t5ddFm/v31ZvbsK+GoXu24fkxf0np3IDUpOuwChAUFY4Ikr7iUVVvzOLZfh5C8MRSX\nlvPNxt18uHI7H63eQXZhKTFREZw8KIVxw7pw6uGpQZuU54uSsgqWZ+SwZPNe4tpGMqRrEoO7JQUl\n6+vGXQXMWraVWd9tZWtOEfFtI5lwRFcuOqoHx/TtEJAhvMWl5bydnsFzX26qnA/TIb4tQ7slMaRb\nEkO6uuHQfTvFN3qOSzBZUDCmFSgrr2Dx5r3MW7mdeat2sD2vmDYRwvEDOjFuaBdOPbxz0D/VlpVX\n8MPWXL7ZuIdFm/aQvjmbohpWkOvZIZahXZMr30iHdEuia3KM32XL3lfC+99nMXPZVpZn5BAhMGZg\nChcf1Z0zh3Qhtm1wmtTKyiv4LiOH1VmemfPb8li3vYCS8goAYqMiObxrYuWcmRE92zEoNbHZBIpm\nERREZBzwLyASeEFVH6n2fDTwCjAK2ANcpqqb6zqnBQVjalZRoazIzOHDVdv5cOV2tuwpBKBtZASp\nydF0SYohNSmGLkkxdEn2/JzsHndOiva5f6K8QvlxWx4LN+5h4aY9LP5pLwX7ywA4LDWR4/p35Nh+\nHTm2XwdKyitYlZVXmXpkdVYem/fsw/u20z4uiiHdkujbKR6AsnKlrEIpr1BKyysorzj48f7SCr7L\nyKa0XDksNZGLR3Xn/BHdSU3yPSdWIJWUVbBxVwGrsg5OsZJf7P4/EqPbcFTv9ozu0560Ph0Y3qNd\n0IJWfUIeFEQkElgHnAFk4tZsnqSqq6vs8wvgSFW9RUQuBy5U1cvqOq8FBWPqp6qs3ZHPwo172J5b\nzPa8YrbnFrMjz/1cXFpxyDFRkUKEuK/ICCFCXGrzyAhBRIj0bM8rLq180+vXKZ5j+3fkeE8g6FTH\nfA6vgv1lrN2e5/nE7d5Ef95bSKQIbSKFNhERREYIbTzXjowQoiIPbDuyRzsuHtW9Qf0ETUFVydhb\nxLKfs1myeS/pm7NZuyMfgDYRwrDuyZVBIq13+zrnwARScwgKxwEPqupZnsf3Aajq/1bZZ55nn4Ui\n0gbYDqRoHYWyoGBM46gqeUVlLlDkFbPDEzSKSsup8Hwqr1CoUPdzuWrl9nJVYqIiGd2nPcf16+RX\n1trWLLewlKU/uwCRvjmb5Zk5lJS5wJyaFE2biAhEcF+I5zuICALgeTzp6F7ceGLDZrX7GhSCOVyh\nO5BR5XEmcExt+6hqmYjkAh2B3UEslzGtmoiQHBdFclyUzY5uIslxUZx6eCqnHp4KuFnwK7fmsmRz\nNht2FlChCu4fqur5fvBjlCaZbR0WY9hEZAowBaBXr14hLo0xxjROdJtIRvXuELD8WIEUzLnyW4Gq\nSeB7eLbVuI+n+SgZ1+F8EFWdqqppqpqWkpISpOIaY4wJZlBYAgwUkb4i0ha4HJhdbZ/ZwDWeny8B\nPq2rP8EYY0xwBa35yNNHcBswDzckdZqqrhKRh4F0VZ0NvAi8KiIbgL24wGGMMSZEgtqnoKpzgDnV\ntv2pys/FwKXBLIMxxhjfWf5dY4wxlSwoGGOMqWRBwRhjTCULCsYYYyqFXZZUEdkFbGng4Z1oebOl\nW9o9tbT7gZZ3Ty3tfqDl3VNN99NbVeud6BV2QaExRCTdl9wf4aSl3VNLux9oeffU0u4HWt49NeZ+\nrPnIGGNMJQsKxhhjKrW2oDA11AUIgpZ2Ty3tfqDl3VNLux9oeffU4PtpVX0Kxhhj6tbaagrGGGPq\nYEHBGGNMpVYTFERknIisFZENInJvqMsTCCKyWUR+EJHlIhJ2a5SKyDQR2SkiK6ts6yAiH4nIes/3\n9qEso79quacHRWSr53VaLiITQllGf4hITxH5TERWi8gqEbnDsz0sX6c67iecX6MYEVksIis89/SQ\nZ3tfEfnW8573lmcJg/rP1xr6FEQkElgHnIFbFnQJMElVV4e0YI0kIpuBNFUNy0k3InISUAC8oqrD\nPNv+CuxV1Uc8wbu9qt4TynL6o5Z7ehAoUNW/h7JsDSEiXYGuqrpMRBKBpcAFwLWE4etUx/1MJHxf\nIwHiVbVARKKAr4A7gN8A76rqdBF5Flihqs/Ud77WUlM4GtigqptUtQSYDpwf4jK1eqr6JW4djarO\nB172/Pwy7g82bNRyT2FLVbep6jLPz/nAj7i11cPydarjfsKWOgWeh1GeLwVOBd7xbPf5NWotQaE7\nkFHlcSZh/ovgocB8EVnqWce6JUhV1W2en7cDqaEsTADdJiLfe5qXwqKppToR6QOMBL6lBbxO1e4H\nwvg1EpFIEVkO7AQ+AjYCOapa5tnF5/e81hIUWqoxqnoUMB74pafposXwLM3aEto3nwH6AyOAbcA/\nQlsc/4lIAjATuFNV86o+F46vUw33E9avkaqWq+oIoAeuZeTwhp6rtQSFrUDPKo97eLaFNVXd6vm+\nE5iF+2UIdzs87b7e9t+dIS5Po6nqDs8fbQXwPGH2OnnaqWcCr6vqu57NYfs61XQ/4f4aealqDvAZ\ncBzQTkS8q2v6/J7XWoLCEmCgpze+LW4t6NkhLlOjiEi8p6MMEYkHzgRW1n1UWJgNXOP5+RrgPyEs\nS0B43zw9LiSMXidPJ+aLwI+q+miVp8LydartfsL8NUoRkXaen2NxA2p+xAWHSzy7+fwatYrRRwCe\nIWaPAZHANFX9S4iL1Cgi0g9XOwC31vYb4XZPIvImMBaX5ncH8ADwHjAD6IVLkT5RVcOm47aWexqL\na5ZQYDNwc5X2+GZNRMYAC4AfgArP5t/j2uHD7nWq434mEb6v0ZG4juRI3Af9Gar6sOc9YjrQAfgO\nmKyq++s9X2sJCsYYY+rXWpqPjDHG+MCCgjHGmEoWFIwxxlSyoGCMMaaSBQVjjDGVLCgY04REZKyI\nvIKtymQAAAGqSURBVB/qchhTGwsKxhhjKllQMKYGIjLZk6N+uYg850k4ViAi//TkrP9ERFI8+44Q\nkUWeZGqzvMnURGSAiHzsyXO/TET6e06fICLviMgaEXndM8vWmGbBgoIx1YjIYOAy4ARPkrFy4Eog\nHkhX1aHAF7jZygCvAPeo6pG4mbLe7a8DT6nqcOB4XKI1cJk57wSGAP2AE4J+U8b4qE39uxjT6pwG\njAKWeD7Ex+ISvlUAb3n2eQ14V0SSgXaq+oVn+8vA2568VN1VdRaAqhYDeM63WFUzPY+XA31wC6MY\nE3IWFIw5lAAvq+p9B20U+WO1/RqaI6Zq/ply7O/QNCPWfGTMoT4BLhGRzlC5HnFv3N+LN+vkFcBX\nqpoLZIvIiZ7tVwFfeFb1yhSRCzzniBaRuCa9C2MawD6hGFONqq4Wkftxq9pFAKXAL4F9wNGe53bi\n+h3ApSV+1vOmvwm4zrP9KuA5EXnYc45Lm/A2jGkQy5JqjI9EpEBVE0JdDmOCyZqPjDHGVLKagjHG\nmEpWUzDGGFPJgoIxxphKFhSMMcZUsqBgjDGmkgUFY4wxlf4/iADi6JEsCpcAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f62340bb978>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# list all data in history\n",
    "print(history1.history.keys())\n",
    "# summarize history for accuracy\n",
    "plt.plot(history1.history['acc'])\n",
    "plt.plot(history1.history['val_acc'])\n",
    "plt.title('model accuracy')\n",
    "plt.ylabel('accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()\n",
    "# summarize history for loss\n",
    "plt.plot(history1.history['loss'])\n",
    "plt.plot(history1.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.ylabel('loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "8424/8424 [==============================] - 1s 141us/step\n"
     ]
    }
   ],
   "source": [
    "# Make predictions\n",
    "prediction = improved_cnn.predict(X_test, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[  8.45739196e-05],\n",
       "       [  9.97828543e-01],\n",
       "       [  9.82402382e-10],\n",
       "       ..., \n",
       "       [  1.23396376e-03],\n",
       "       [  9.99999881e-01],\n",
       "       [  2.27747127e-01]], dtype=float32)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}