finalProject-res50-part-train.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Technologies\n",
    "# CUDA toolkit 10.0\n",
    "# Tensorflow-gpu 2.0\n",
    "# cuDNN 7.6.4"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from scipy.io import loadmat\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from math import ceil\n",
    "import tensorflow as tf\n",
    "from PIL import Image, ImageFile"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "VirtualDeviceConfiguration(memory_limit=6040)"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ImageFile.LOAD_TRUNCATED_IMAGES = True\n",
    "tf.config.experimental.VirtualDeviceConfiguration(memory_limit=6040)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "img nummber:  16185\n",
      "label number:  196\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>imgPath</th>\n",
       "      <th>label</th>\n",
       "      <th>testTag</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>16175</th>\n",
       "      <td>car_ims/016176.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16176</th>\n",
       "      <td>car_ims/016177.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16177</th>\n",
       "      <td>car_ims/016178.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16178</th>\n",
       "      <td>car_ims/016179.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16179</th>\n",
       "      <td>car_ims/016180.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16180</th>\n",
       "      <td>car_ims/016181.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16181</th>\n",
       "      <td>car_ims/016182.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16182</th>\n",
       "      <td>car_ims/016183.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16183</th>\n",
       "      <td>car_ims/016184.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16184</th>\n",
       "      <td>car_ims/016185.jpg</td>\n",
       "      <td>smart fortwo Convertible 2012</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                  imgPath                          label  testTag\n",
       "16175  car_ims/016176.jpg  smart fortwo Convertible 2012        1\n",
       "16176  car_ims/016177.jpg  smart fortwo Convertible 2012        1\n",
       "16177  car_ims/016178.jpg  smart fortwo Convertible 2012        1\n",
       "16178  car_ims/016179.jpg  smart fortwo Convertible 2012        1\n",
       "16179  car_ims/016180.jpg  smart fortwo Convertible 2012        1\n",
       "16180  car_ims/016181.jpg  smart fortwo Convertible 2012        1\n",
       "16181  car_ims/016182.jpg  smart fortwo Convertible 2012        1\n",
       "16182  car_ims/016183.jpg  smart fortwo Convertible 2012        1\n",
       "16183  car_ims/016184.jpg  smart fortwo Convertible 2012        1\n",
       "16184  car_ims/016185.jpg  smart fortwo Convertible 2012        1"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#map betweem image names and labels \n",
    "matData = loadmat('./cars_annos.mat')\n",
    "\n",
    "labelsList = []\n",
    "for arr in matData['annotations'][0]:\n",
    "    imgPath = arr[0][0]\n",
    "    labelNum = arr[5][0][0]\n",
    "    splitTag = arr[6][0][0]\n",
    "    labelsList.append([imgPath, labelNum, splitTag])\n",
    "print(\"img nummber: \", len(labelsList))\n",
    "\n",
    "labelNameDict = {}\n",
    "labelList = []\n",
    "for i,arr in enumerate(matData['class_names'][0]):\n",
    "    labelName = arr[0]\n",
    "    labelNameDict[i+1] = labelName\n",
    "    labelList.append(labelName)\n",
    "print(\"label number: \", len(labelNameDict))\n",
    "\n",
    "labelsDF = pd.DataFrame(labelsList, columns=['imgPath', 'label', 'testTag'])\n",
    "labelsDF['label'] = labelsDF['label'].map(labelNameDict)\n",
    "labelsDF.tail(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "amount of train set: 8144\n",
      "amount of test set: 8041\n",
      "amount of validationDF: 1000\n"
     ]
    }
   ],
   "source": [
    "#split train and test\n",
    "trainDF = labelsDF[labelsDF['testTag']==0]\n",
    "print(\"amount of train set:\", len(trainDF))\n",
    "testDF = labelsDF[labelsDF['testTag']==1]\n",
    "print(\"amount of test set:\", len(testDF))\n",
    "validationDF = testDF.sample(n=1000)\n",
    "print(\"amount of validationDF:\", len(validationDF))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model parms\n",
    "classNum = len(labelNameDict)\n",
    "classContent = labelList"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "#base model\n",
    "base_model = tf.keras.applications.resnet.ResNet50(weights='resnet50_weights_tf_dim_ordering_tf_kernels_notop.h5', include_top=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#add top layer\n",
    "top = base_model.output\n",
    "top = tf.keras.layers.GlobalAveragePooling2D()(top)\n",
    "top = tf.keras.layers.Flatten()(top)\n",
    "top = tf.keras.layers.Dense(2048, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.0001))(top)\n",
    "top = tf.keras.layers.BatchNormalization()(top)\n",
    "top = tf.keras.layers.Dense(1024, activation='relu', kernel_regularizer=tf.keras.regularizers.l2(0.0001))(top)\n",
    "top = tf.keras.layers.BatchNormalization(name='bn_fc_01')(top)\n",
    "top_model = tf.keras.layers.Dense(classNum, activation='softmax')(top)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# learning rate\n",
    "# adam = tf.keras.optimizers.Adam(lr=0.001, beta_1=0.9, beta_2=0.999, epsilon=1e-08)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#create model\n",
    "model = tf.keras.models.Model(inputs=base_model.input, outputs=top_model)\n",
    "for layer in base_model.layers[:30]:\n",
    "    layer.trainable = False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model load\n",
    "#model = tf.keras.models.load_model('models/3/car.15-0.17.hdf5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model config\n",
    "adam = tf.keras.optimizers.Adam(learning_rate=0.01)\n",
    "model.compile(optimizer= adam, loss='categorical_crossentropy', metrics=['acc'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "#call back functions\n",
    "early_stop = tf.keras.callbacks.EarlyStopping('loss', patience=5)\n",
    "model_checkpoint = tf.keras.callbacks.ModelCheckpoint('./models/car.{epoch:02d}-{val_acc:.2f}.hdf5', monitor='val_acc', verbose=1, save_best_only=True)\n",
    "reduce_lr = tf.keras.callbacks.ReduceLROnPlateau('loss', factor=0.1, patience=1)\n",
    "callbacks = [early_stop, model_checkpoint, reduce_lr]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "#data preprocess parms\n",
    "img_width, img_height = 224, 224\n",
    "batch_size = 10\n",
    "epochs = 100\n",
    "trainNum = len(trainDF)\n",
    "validNum = len(validationDF)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 8144 validated image filenames belonging to 196 classes.\n"
     ]
    }
   ],
   "source": [
    "#input data preprocess\n",
    "train_datagen = tf.keras.preprocessing.image.ImageDataGenerator(\n",
    "    rotation_range = 20,\n",
    "    width_shift_range = 0.1,\n",
    "    height_shift_range = 0.1,\n",
    "    zoom_range = 0.2,\n",
    "    horizontal_flip = True\n",
    ")\n",
    "train_generator = train_datagen.flow_from_dataframe(\n",
    "    #trainDF.sample(n=trainNum),\n",
    "    trainDF,\n",
    "    target_size = (img_width, img_height),\n",
    "    x_col = 'imgPath',\n",
    "    y_col = 'label',\n",
    "    classes = classContent,\n",
    "    batch_size = batch_size,\n",
    "    class_mode = 'categorical'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 1000 validated image filenames belonging to 196 classes.\n"
     ]
    }
   ],
   "source": [
    "#valid data preprocess\n",
    "valid_datagen = tf.keras.preprocessing.image.ImageDataGenerator()\n",
    "valid_generator = valid_datagen.flow_from_dataframe(\n",
    "    validationDF,\n",
    "    target_size = (img_width, img_height),\n",
    "    x_col = 'imgPath',\n",
    "    y_col = 'label',\n",
    "    classes = classContent,\n",
    "    batch_size = batch_size,\n",
    "    class_mode = 'categorical'\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 8.5001 - acc: 0.0045\n",
      "Epoch 00001: val_acc improved from -inf to 0.00400, saving model to ./models/car.01-0.00.hdf5\n",
      "815/815 [==============================] - 322s 395ms/step - loss: 8.4976 - acc: 0.0045 - val_loss: 7.8909 - val_acc: 0.0040\n",
      "Epoch 2/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 6.3594 - acc: 0.0053\n",
      "Epoch 00002: val_acc improved from 0.00400 to 0.01000, saving model to ./models/car.02-0.01.hdf5\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 6.3593 - acc: 0.0053 - val_loss: 8.7717 - val_acc: 0.0100\n",
      "Epoch 3/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 6.2980 - acc: 0.0047\n",
      "Epoch 00003: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 6.2978 - acc: 0.0047 - val_loss: 7.4588 - val_acc: 0.0050\n",
      "Epoch 4/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 6.2094 - acc: 0.0052\n",
      "Epoch 00004: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 299s 367ms/step - loss: 6.2091 - acc: 0.0052 - val_loss: 6.7165 - val_acc: 0.0040\n",
      "Epoch 5/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 6.0610 - acc: 0.0069\n",
      "Epoch 00005: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 299s 367ms/step - loss: 6.0612 - acc: 0.0069 - val_loss: 15.5601 - val_acc: 0.0030\n",
      "Epoch 6/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 6.1607 - acc: 0.0050\n",
      "Epoch 00006: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 306s 375ms/step - loss: 6.1602 - acc: 0.0050 - val_loss: 10.0415 - val_acc: 0.0080\n",
      "Epoch 7/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.7323 - acc: 0.0061\n",
      "Epoch 00007: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 5.7322 - acc: 0.0061 - val_loss: 5.8505 - val_acc: 0.0060\n",
      "Epoch 8/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.5500 - acc: 0.0073\n",
      "Epoch 00008: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 301s 370ms/step - loss: 5.5499 - acc: 0.0072 - val_loss: 5.7588 - val_acc: 0.0060\n",
      "Epoch 9/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.4296 - acc: 0.0082\n",
      "Epoch 00009: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 300s 368ms/step - loss: 5.4293 - acc: 0.0082 - val_loss: 6.4262 - val_acc: 0.0070\n",
      "Epoch 10/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.3512 - acc: 0.0091\n",
      "Epoch 00010: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 5.3513 - acc: 0.0091 - val_loss: 5.8660 - val_acc: 0.0080\n",
      "Epoch 11/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.2877 - acc: 0.0120\n",
      "Epoch 00011: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 299s 367ms/step - loss: 5.2875 - acc: 0.0120 - val_loss: 5.7088 - val_acc: 0.0040\n",
      "Epoch 12/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.2371 - acc: 0.0130\n",
      "Epoch 00012: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 5.2370 - acc: 0.0130 - val_loss: 5.3593 - val_acc: 0.0070\n",
      "Epoch 13/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.1970 - acc: 0.0134\n",
      "Epoch 00013: val_acc did not improve from 0.01000\n",
      "815/815 [==============================] - 307s 376ms/step - loss: 5.1968 - acc: 0.0134 - val_loss: 5.3613 - val_acc: 0.0070\n",
      "Epoch 14/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.1400 - acc: 0.0141\n",
      "Epoch 00014: val_acc improved from 0.01000 to 0.01100, saving model to ./models/car.14-0.01.hdf5\n",
      "815/815 [==============================] - 303s 372ms/step - loss: 5.1401 - acc: 0.0141 - val_loss: 6.0933 - val_acc: 0.0110\n",
      "Epoch 15/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 5.0518 - acc: 0.0149\n",
      "Epoch 00015: val_acc improved from 0.01100 to 0.01600, saving model to ./models/car.15-0.02.hdf5\n",
      "815/815 [==============================] - 301s 370ms/step - loss: 5.0521 - acc: 0.0149 - val_loss: 5.1685 - val_acc: 0.0160\n",
      "Epoch 16/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.9301 - acc: 0.0205\n",
      "Epoch 00016: val_acc did not improve from 0.01600\n",
      "815/815 [==============================] - 305s 375ms/step - loss: 4.9299 - acc: 0.0205 - val_loss: 5.4941 - val_acc: 0.0130\n",
      "Epoch 17/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.8451 - acc: 0.0245\n",
      "Epoch 00017: val_acc did not improve from 0.01600\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 4.8454 - acc: 0.0246 - val_loss: 5.6167 - val_acc: 0.0130\n",
      "Epoch 18/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.7638 - acc: 0.0290\n",
      "Epoch 00018: val_acc did not improve from 0.01600\n",
      "815/815 [==============================] - 300s 368ms/step - loss: 4.7639 - acc: 0.0291 - val_loss: 5.4086 - val_acc: 0.0130\n",
      "Epoch 19/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.6677 - acc: 0.0349\n",
      "Epoch 00019: val_acc did not improve from 0.01600\n",
      "815/815 [==============================] - 306s 376ms/step - loss: 4.6675 - acc: 0.0349 - val_loss: 9.6254 - val_acc: 0.0060\n",
      "Epoch 20/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.5715 - acc: 0.0416\n",
      "Epoch 00020: val_acc did not improve from 0.01600\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 4.5716 - acc: 0.0415 - val_loss: 8.0780 - val_acc: 0.0060\n",
      "Epoch 21/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.4555 - acc: 0.0521\n",
      "Epoch 00021: val_acc improved from 0.01600 to 0.02800, saving model to ./models/car.21-0.03.hdf5\n",
      "815/815 [==============================] - 301s 370ms/step - loss: 4.4553 - acc: 0.0521 - val_loss: 5.9960 - val_acc: 0.0280\n",
      "Epoch 22/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.3386 - acc: 0.0610\n",
      "Epoch 00022: val_acc improved from 0.02800 to 0.04500, saving model to ./models/car.22-0.05.hdf5\n",
      "815/815 [==============================] - 308s 377ms/step - loss: 4.3385 - acc: 0.0609 - val_loss: 5.3528 - val_acc: 0.0450\n",
      "Epoch 23/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.2393 - acc: 0.0701\n",
      "Epoch 00023: val_acc did not improve from 0.04500\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 4.2393 - acc: 0.0700 - val_loss: 5.5981 - val_acc: 0.0440\n",
      "Epoch 24/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 4.1076 - acc: 0.0851\n",
      "Epoch 00024: val_acc did not improve from 0.04500\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 4.1072 - acc: 0.0850 - val_loss: 6.0539 - val_acc: 0.0210\n",
      "Epoch 25/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.9636 - acc: 0.0956\n",
      "Epoch 00025: val_acc did not improve from 0.04500\n",
      "815/815 [==============================] - 301s 370ms/step - loss: 3.9637 - acc: 0.0955 - val_loss: 5.7019 - val_acc: 0.0420\n",
      "Epoch 26/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.8113 - acc: 0.1113\n",
      "Epoch 00026: val_acc did not improve from 0.04500\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 3.8106 - acc: 0.1112 - val_loss: 7.3853 - val_acc: 0.0300\n",
      "Epoch 27/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.6606 - acc: 0.1404\n",
      "Epoch 00027: val_acc did not improve from 0.04500\n",
      "815/815 [==============================] - 302s 370ms/step - loss: 3.6607 - acc: 0.1403 - val_loss: 6.1475 - val_acc: 0.0410\n",
      "Epoch 28/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.4711 - acc: 0.1646\n",
      "Epoch 00028: val_acc did not improve from 0.04500\n",
      "815/815 [==============================] - 302s 370ms/step - loss: 3.4718 - acc: 0.1644 - val_loss: 13.4208 - val_acc: 0.0130\n",
      "Epoch 29/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.3596 - acc: 0.1815\n",
      "Epoch 00029: val_acc improved from 0.04500 to 0.10600, saving model to ./models/car.29-0.11.hdf5\n",
      "815/815 [==============================] - 309s 379ms/step - loss: 3.3598 - acc: 0.1815 - val_loss: 4.5473 - val_acc: 0.1060\n",
      "Epoch 30/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.2205 - acc: 0.2140\n",
      "Epoch 00030: val_acc did not improve from 0.10600\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 3.2199 - acc: 0.2140 - val_loss: 5.6509 - val_acc: 0.0780\n",
      "Epoch 31/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 3.0440 - acc: 0.2408\n",
      "Epoch 00031: val_acc did not improve from 0.10600\n",
      "815/815 [==============================] - 299s 367ms/step - loss: 3.0425 - acc: 0.2412 - val_loss: 9.0543 - val_acc: 0.0500\n",
      "Epoch 32/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.9081 - acc: 0.2719\n",
      "Epoch 00032: val_acc did not improve from 0.10600\n",
      "815/815 [==============================] - 300s 369ms/step - loss: 2.9086 - acc: 0.2720 - val_loss: 4.8591 - val_acc: 0.0870\n",
      "Epoch 33/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.7791 - acc: 0.3006\n",
      "Epoch 00033: val_acc did not improve from 0.10600\n",
      "815/815 [==============================] - 303s 371ms/step - loss: 2.7794 - acc: 0.3002 - val_loss: 9.1863 - val_acc: 0.0510\n",
      "Epoch 34/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.6195 - acc: 0.3417\n",
      "Epoch 00034: val_acc did not improve from 0.10600\n",
      "815/815 [==============================] - 299s 367ms/step - loss: 2.6192 - acc: 0.3416 - val_loss: 8.6933 - val_acc: 0.0550\n",
      "Epoch 35/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.5081 - acc: 0.3624\n",
      "Epoch 00035: val_acc did not improve from 0.10600\n",
      "815/815 [==============================] - 300s 368ms/step - loss: 2.5088 - acc: 0.3622 - val_loss: 8.8292 - val_acc: 0.0580\n",
      "Epoch 36/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.3665 - acc: 0.3912\n",
      "Epoch 00036: val_acc improved from 0.10600 to 0.11800, saving model to ./models/car.36-0.12.hdf5\n",
      "815/815 [==============================] - 308s 378ms/step - loss: 2.3664 - acc: 0.3913 - val_loss: 5.6437 - val_acc: 0.1180\n",
      "Epoch 37/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.2952 - acc: 0.4053\n",
      "Epoch 00037: val_acc improved from 0.11800 to 0.17800, saving model to ./models/car.37-0.18.hdf5\n",
      "815/815 [==============================] - 304s 372ms/step - loss: 2.2958 - acc: 0.4055 - val_loss: 4.1927 - val_acc: 0.1780\n",
      "Epoch 38/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.1837 - acc: 0.4336\n",
      "Epoch 00038: val_acc did not improve from 0.17800\n",
      "815/815 [==============================] - 300s 368ms/step - loss: 2.1836 - acc: 0.4334 - val_loss: 12.4875 - val_acc: 0.0450\n",
      "Epoch 39/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 2.0542 - acc: 0.4646\n",
      "Epoch 00039: val_acc did not improve from 0.17800\n",
      "815/815 [==============================] - 306s 376ms/step - loss: 2.0533 - acc: 0.4648 - val_loss: 14.0186 - val_acc: 0.0320\n",
      "Epoch 40/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.9779 - acc: 0.4864\n",
      "Epoch 00040: val_acc did not improve from 0.17800\n",
      "815/815 [==============================] - 302s 370ms/step - loss: 1.9768 - acc: 0.4869 - val_loss: 12.7357 - val_acc: 0.0370\n",
      "Epoch 41/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.8827 - acc: 0.5175\n",
      "Epoch 00041: val_acc did not improve from 0.17800\n",
      "815/815 [==============================] - 300s 368ms/step - loss: 1.8830 - acc: 0.5173 - val_loss: 8.2685 - val_acc: 0.0820\n",
      "Epoch 42/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.7944 - acc: 0.5298\n",
      "Epoch 00042: val_acc did not improve from 0.17800\n",
      "815/815 [==============================] - 299s 367ms/step - loss: 1.7950 - acc: 0.5292 - val_loss: 5.1794 - val_acc: 0.1740\n",
      "Epoch 43/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.7110 - acc: 0.5460\n",
      "Epoch 00043: val_acc improved from 0.17800 to 0.20800, saving model to ./models/car.43-0.21.hdf5\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 1.7103 - acc: 0.5462 - val_loss: 4.3687 - val_acc: 0.2080\n",
      "Epoch 44/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.6661 - acc: 0.5604\n",
      "Epoch 00044: val_acc did not improve from 0.20800\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 1.6668 - acc: 0.5604 - val_loss: 6.6593 - val_acc: 0.1080\n",
      "Epoch 45/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.5861 - acc: 0.5819\n",
      "Epoch 00045: val_acc did not improve from 0.20800\n",
      "815/815 [==============================] - 303s 371ms/step - loss: 1.5852 - acc: 0.5821 - val_loss: 13.6049 - val_acc: 0.0310\n",
      "Epoch 46/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.5202 - acc: 0.6003\n",
      "Epoch 00046: val_acc did not improve from 0.20800\n",
      "815/815 [==============================] - 306s 375ms/step - loss: 1.5208 - acc: 0.6002 - val_loss: 12.1718 - val_acc: 0.0680\n",
      "Epoch 47/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.4623 - acc: 0.6170\n",
      "Epoch 00047: val_acc did not improve from 0.20800\n",
      "815/815 [==============================] - 302s 370ms/step - loss: 1.4626 - acc: 0.6169 - val_loss: 11.0864 - val_acc: 0.0640\n",
      "Epoch 48/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.4188 - acc: 0.6298\n",
      "Epoch 00048: val_acc did not improve from 0.20800\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 1.4191 - acc: 0.6298 - val_loss: 9.2383 - val_acc: 0.1010\n",
      "Epoch 49/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.3287 - acc: 0.6499\n",
      "Epoch 00049: val_acc improved from 0.20800 to 0.22300, saving model to ./models/car.49-0.22.hdf5\n",
      "815/815 [==============================] - 303s 371ms/step - loss: 1.3299 - acc: 0.6494 - val_loss: 4.6213 - val_acc: 0.2230\n",
      "Epoch 50/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.3007 - acc: 0.6590\n",
      "Epoch 00050: val_acc did not improve from 0.22300\n",
      "815/815 [==============================] - 308s 378ms/step - loss: 1.3007 - acc: 0.6588 - val_loss: 5.1134 - val_acc: 0.2080\n",
      "Epoch 51/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.2342 - acc: 0.6749\n",
      "Epoch 00051: val_acc did not improve from 0.22300\n",
      "815/815 [==============================] - 303s 371ms/step - loss: 1.2341 - acc: 0.6750 - val_loss: 10.5448 - val_acc: 0.0660\n",
      "Epoch 52/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.2146 - acc: 0.6821\n",
      "Epoch 00052: val_acc improved from 0.22300 to 0.23500, saving model to ./models/car.52-0.23.hdf5\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 1.2145 - acc: 0.6822 - val_loss: 4.6771 - val_acc: 0.2350\n",
      "Epoch 53/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.1804 - acc: 0.6902\n",
      "Epoch 00053: val_acc improved from 0.23500 to 0.24100, saving model to ./models/car.53-0.24.hdf5\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 1.1797 - acc: 0.6904 - val_loss: 4.8024 - val_acc: 0.2410\n",
      "Epoch 54/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.1447 - acc: 0.6993\n",
      "Epoch 00054: val_acc did not improve from 0.24100\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 1.1456 - acc: 0.6992 - val_loss: 9.6662 - val_acc: 0.0690\n",
      "Epoch 55/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.1106 - acc: 0.7086\n",
      "Epoch 00055: val_acc did not improve from 0.24100\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 1.1101 - acc: 0.7087 - val_loss: 5.7197 - val_acc: 0.1910\n",
      "Epoch 56/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.0774 - acc: 0.7241\n",
      "Epoch 00056: val_acc did not improve from 0.24100\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 1.0773 - acc: 0.7240 - val_loss: 12.9991 - val_acc: 0.0550\n",
      "Epoch 57/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.0222 - acc: 0.7360\n",
      "Epoch 00057: val_acc improved from 0.24100 to 0.28600, saving model to ./models/car.57-0.29.hdf5\n",
      "815/815 [==============================] - 305s 374ms/step - loss: 1.0226 - acc: 0.7358 - val_loss: 4.9004 - val_acc: 0.2860\n",
      "Epoch 58/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 1.0139 - acc: 0.7353\n",
      "Epoch 00058: val_acc did not improve from 0.28600\n",
      "815/815 [==============================] - 300s 368ms/step - loss: 1.0142 - acc: 0.7353 - val_loss: 6.1515 - val_acc: 0.1900\n",
      "Epoch 59/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.9901 - acc: 0.7421\n",
      "Epoch 00059: val_acc did not improve from 0.28600\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 0.9895 - acc: 0.7424 - val_loss: 8.8169 - val_acc: 0.1060\n",
      "Epoch 60/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.9299 - acc: 0.7628\n",
      "Epoch 00060: val_acc did not improve from 0.28600\n",
      "815/815 [==============================] - 304s 374ms/step - loss: 0.9299 - acc: 0.7629 - val_loss: 5.9565 - val_acc: 0.2270\n",
      "Epoch 61/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.9418 - acc: 0.7553\n",
      "Epoch 00061: val_acc did not improve from 0.28600\n",
      "815/815 [==============================] - 304s 374ms/step - loss: 0.9414 - acc: 0.7554 - val_loss: 11.2701 - val_acc: 0.0630\n",
      "Epoch 62/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.7133 - acc: 0.8269\n",
      "Epoch 00062: val_acc improved from 0.28600 to 0.52700, saving model to ./models/car.62-0.53.hdf5\n",
      "815/815 [==============================] - 303s 372ms/step - loss: 0.7130 - acc: 0.8270 - val_loss: 2.1922 - val_acc: 0.5270\n",
      "Epoch 63/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.6281 - acc: 0.8555\n",
      "Epoch 00063: val_acc did not improve from 0.52700\n",
      "815/815 [==============================] - 302s 371ms/step - loss: 0.6278 - acc: 0.8556 - val_loss: 2.4277 - val_acc: 0.5100\n",
      "Epoch 64/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.5818 - acc: 0.8653\n",
      "Epoch 00064: val_acc improved from 0.52700 to 0.54100, saving model to ./models/car.64-0.54.hdf5\n",
      "815/815 [==============================] - 304s 373ms/step - loss: 0.5823 - acc: 0.8653 - val_loss: 2.2188 - val_acc: 0.5410\n",
      "Epoch 65/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.5386 - acc: 0.8736\n",
      "Epoch 00065: val_acc did not improve from 0.54100\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.5387 - acc: 0.8734 - val_loss: 2.2532 - val_acc: 0.5410\n",
      "Epoch 66/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.5304 - acc: 0.8769\n",
      "Epoch 00066: val_acc did not improve from 0.54100\n",
      "815/815 [==============================] - 303s 372ms/step - loss: 0.5313 - acc: 0.8766 - val_loss: 2.3188 - val_acc: 0.5330\n",
      "Epoch 67/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.5039 - acc: 0.8823\n",
      "Epoch 00067: val_acc improved from 0.54100 to 0.55500, saving model to ./models/car.67-0.56.hdf5\n",
      "815/815 [==============================] - 304s 373ms/step - loss: 0.5039 - acc: 0.8824 - val_loss: 2.2101 - val_acc: 0.5550\n",
      "Epoch 68/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4893 - acc: 0.8907\n",
      "Epoch 00068: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.4890 - acc: 0.8908 - val_loss: 2.3096 - val_acc: 0.5340\n",
      "Epoch 69/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4764 - acc: 0.8908\n",
      "Epoch 00069: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 304s 372ms/step - loss: 0.4765 - acc: 0.8906 - val_loss: 2.2970 - val_acc: 0.5290\n",
      "Epoch 70/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4717 - acc: 0.8940\n",
      "Epoch 00070: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 303s 372ms/step - loss: 0.4716 - acc: 0.8940 - val_loss: 2.6108 - val_acc: 0.5200\n",
      "Epoch 71/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4592 - acc: 0.8944\n",
      "Epoch 00071: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.4591 - acc: 0.8943 - val_loss: 2.7454 - val_acc: 0.4830\n",
      "Epoch 72/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4488 - acc: 0.8976\n",
      "Epoch 00072: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 0.4488 - acc: 0.8975 - val_loss: 2.3663 - val_acc: 0.5370\n",
      "Epoch 73/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4301 - acc: 0.9026\n",
      "Epoch 00073: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 303s 372ms/step - loss: 0.4300 - acc: 0.9026 - val_loss: 2.2164 - val_acc: 0.5430\n",
      "Epoch 74/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4334 - acc: 0.9013\n",
      "Epoch 00074: val_acc did not improve from 0.55500\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.4338 - acc: 0.9012 - val_loss: 2.2989 - val_acc: 0.5470\n",
      "Epoch 75/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4197 - acc: 0.9062\n",
      "Epoch 00075: val_acc improved from 0.55500 to 0.56200, saving model to ./models/car.75-0.56.hdf5\n",
      "815/815 [==============================] - 308s 378ms/step - loss: 0.4195 - acc: 0.9063 - val_loss: 2.2572 - val_acc: 0.5620\n",
      "Epoch 76/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.4162 - acc: 0.9031\n",
      "Epoch 00076: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 303s 371ms/step - loss: 0.4163 - acc: 0.9031 - val_loss: 2.2554 - val_acc: 0.5570\n",
      "Epoch 77/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3931 - acc: 0.9068\n",
      "Epoch 00077: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.3935 - acc: 0.9067 - val_loss: 2.2696 - val_acc: 0.5480\n",
      "Epoch 78/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3984 - acc: 0.9139\n",
      "Epoch 00078: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.3982 - acc: 0.9140 - val_loss: 2.3179 - val_acc: 0.5510\n",
      "Epoch 79/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3908 - acc: 0.9109\n",
      "Epoch 00079: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 307s 376ms/step - loss: 0.3905 - acc: 0.9110 - val_loss: 2.3054 - val_acc: 0.5490\n",
      "Epoch 80/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3930 - acc: 0.9114\n",
      "Epoch 00080: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 302s 370ms/step - loss: 0.3926 - acc: 0.9115 - val_loss: 2.2963 - val_acc: 0.5550\n",
      "Epoch 81/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3838 - acc: 0.9155\n",
      "Epoch 00081: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.3836 - acc: 0.9156 - val_loss: 2.3101 - val_acc: 0.5460\n",
      "Epoch 82/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3935 - acc: 0.9111\n",
      "Epoch 00082: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 307s 377ms/step - loss: 0.3937 - acc: 0.9111 - val_loss: 2.2784 - val_acc: 0.5500\n",
      "Epoch 83/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3943 - acc: 0.9121\n",
      "Epoch 00083: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 302s 370ms/step - loss: 0.3940 - acc: 0.9122 - val_loss: 2.2538 - val_acc: 0.5560\n",
      "Epoch 84/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3878 - acc: 0.9139\n",
      "Epoch 00084: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 301s 369ms/step - loss: 0.3880 - acc: 0.9138 - val_loss: 2.2991 - val_acc: 0.5460\n",
      "Epoch 85/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3939 - acc: 0.9096\n",
      "Epoch 00085: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 307s 376ms/step - loss: 0.3939 - acc: 0.9097 - val_loss: 2.2746 - val_acc: 0.5490\n",
      "Epoch 86/100\n",
      "814/815 [============================>.] - ETA: 0s - loss: 0.3987 - acc: 0.9100\n",
      "Epoch 00086: val_acc did not improve from 0.56200\n",
      "815/815 [==============================] - 303s 372ms/step - loss: 0.3984 - acc: 0.9101 - val_loss: 2.3196 - val_acc: 0.5410\n"
     ]
    }
   ],
   "source": [
    "#train model\n",
    "model_history = model.fit_generator(\n",
    "    train_generator,\n",
    "    steps_per_epoch = ceil(trainNum / batch_size),\n",
    "    validation_data = valid_generator,\n",
    "    validation_steps = ceil(validNum / batch_size),\n",
    "    epochs = epochs,\n",
    "    callbacks = callbacks,\n",
    "    verbose=1\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1ceb63d6ec8>"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(0)\n",
    "plt.plot(model_history.history['acc'],'r')\n",
    "plt.plot(model_history.history['val_acc'],'g')\n",
    "plt.xlabel(\"epochs\")\n",
    "plt.ylabel(\"accuracy\")\n",
    "plt.title(\"Training Acc vs Validation Acc\")\n",
    "plt.legend([\"train\", \"validation\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x1ceb6449d88>"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(1)\n",
    "plt.plot(model_history.history['loss'],'r')\n",
    "plt.plot(model_history.history['val_loss'],'g')\n",
    "plt.xlabel(\"epochs\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.title(\"Training Loss vs Validation Loss\")\n",
    "plt.legend([\"train\", \"validation\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tensorflow2",
   "language": "python",
   "name": "tensorflow2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}