disease-classification-by-cnn-using-mfcc.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "_cell_guid": "79c7e3d0-c299-4dcb-8224-4455121ee9b0",
    "_uuid": "d629ff2d2480ee46fbb7e2d37f6b5fab8052498a"
   },
   "outputs": [],
   "source": [
    "# Load various imports \n",
    "from datetime import datetime\n",
    "from os import listdir\n",
    "from os.path import isfile, join\n",
    "\n",
    "import librosa\n",
    "import librosa.display\n",
    "\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense, Dropout, Conv2D, MaxPooling2D, GlobalAveragePooling2D\n",
    "#from keras.utils import to_categorical\n",
    "from tensorflow.keras.utils import to_categorical\n",
    "from tensorflow.keras.callbacks import ModelCheckpoint\n",
    "\n",
    "from sklearn.metrics import confusion_matrix, classification_report, roc_curve, auc\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "mypath = \"./respiratory-sound-database/Respiratory_Sound_Database/Respiratory_Sound_Database/audio_and_txt_files/\"\n",
    "filenames = [f for f in listdir(mypath) if (isfile(join(mypath, f)) and f.endswith('.wav'))] "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "p_id_in_file = [] # patient IDs corresponding to each file\n",
    "for name in filenames:\n",
    "    p_id_in_file.append(int(name[:3]))\n",
    "\n",
    "p_id_in_file = np.array(p_id_in_file) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "max_pad_len = 862 # to make the length of all MFCC equal\n",
    "\n",
    "def extract_features(file_name):\n",
    "    \"\"\"\n",
    "    This function takes in the path for an audio file as a string, loads it, and returns the MFCC\n",
    "    of the audio\"\"\"\n",
    "   \n",
    "    try:\n",
    "        audio, sample_rate = librosa.load(file_name, res_type='kaiser_fast', duration=20) \n",
    "        mfccs = librosa.feature.mfcc(y=audio, sr=sample_rate, n_mfcc=40)\n",
    "        pad_width = max_pad_len - mfccs.shape[1]\n",
    "        mfccs = np.pad(mfccs, pad_width=((0, 0), (0, pad_width)), mode='constant')\n",
    "        \n",
    "    except Exception as e:\n",
    "        print(\"Error encountered while parsing file: \", file_name)\n",
    "        return None \n",
    "     \n",
    "    return mfccs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "filepaths = [join(mypath, f) for f in filenames] # full paths of files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "p_diag = pd.read_csv(\"./respiratory-sound-database/Respiratory_Sound_Database/Respiratory_Sound_Database/patient_diagnosis.csv\",header=None) # patient diagnosis file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = np.array([p_diag[p_diag[0] == x][1].values[0] for x in p_id_in_file]) # labels for audio files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Finished feature extraction from  920  files\n"
     ]
    }
   ],
   "source": [
    "features = [] \n",
    "\n",
    "# Iterate through each sound file and extract the features\n",
    "for file_name in filepaths:\n",
    "    data = extract_features(file_name)\n",
    "    features.append(data)\n",
    "\n",
    "print('Finished feature extraction from ', len(features), ' files')\n",
    "features = np.array(features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot an MFCC\n",
    "plt.figure(figsize=(10, 4))\n",
    "librosa.display.specshow(features[7], x_axis='time')\n",
    "plt.colorbar()\n",
    "plt.title('MFCC')\n",
    "plt.tight_layout()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "features = np.array(features) # convert to numpy array"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# delete the very rare diseases\n",
    "features1 = np.delete(features, np.where((labels == 'Asthma') | (labels == 'LRTI'))[0], axis=0) \n",
    "\n",
    "labels1 = np.delete(labels, np.where((labels == 'Asthma') | (labels == 'LRTI'))[0], axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['Bronchiectasis' 'Bronchiolitis' 'COPD' 'Healthy' 'Pneumonia' 'URTI']\n",
      " ['16' '13' '793' '35' '37' '23']]\n"
     ]
    }
   ],
   "source": [
    "# print class counts\n",
    "unique_elements, counts_elements = np.unique(labels1, return_counts=True)\n",
    "print(np.asarray((unique_elements, counts_elements)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot class counts\n",
    "y_pos = np.arange(len(unique_elements))\n",
    "plt.figure(figsize=(12,8))\n",
    "plt.bar(unique_elements, counts_elements, align='center', alpha=0.5)\n",
    "plt.xticks(y_pos, unique_elements)\n",
    "plt.ylabel('Count')\n",
    "plt.xlabel('Disease')\n",
    "plt.title('Disease Count in Sound Files (No Asthma or LRTI)')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# One-hot encode labels\n",
    "le = LabelEncoder()\n",
    "i_labels = le.fit_transform(labels1)\n",
    "oh_labels = to_categorical(i_labels) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "# add channel dimension for CNN\n",
    "features1 = np.reshape(features1, (*features1.shape,1)) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# train test split\n",
    "x_train, x_test, y_train, y_test = train_test_split(features1, oh_labels, stratify=oh_labels, \n",
    "                                                    test_size=0.2, random_state = 42)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Convolutional Neural Network (CNN) model architecture**\n",
    "\n",
    "The model will be a Convolutional Neural Network (CNN) using Keras and a Tensorflow backend.\n",
    "\n",
    "A sequential model is used, with a simple model architecture, consisting of four Conv2D convolution layers, with the final output layer being a dense layer.\n",
    "\n",
    "The first layer will receive the input shape of (40, 862, 1) where 40 is the number of MFCC's, 862 is the number of frames taking padding into account and the 1 signifying that the audio is mono.\n",
    "\n",
    "A small Dropout value of 20% is used on the convolutional layers.\n",
    "\n",
    "Each convolutional layer has an associated pooling layer of MaxPooling2D type with the final convolutional layer having a GlobalAveragePooling2D type.\n",
    "\n",
    "The output layer will has 6 nodes (num_labels) which matches the number of possible classifications."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "num_rows = 40\n",
    "num_columns = 862\n",
    "num_channels = 1\n",
    "\n",
    "num_labels = oh_labels.shape[1]\n",
    "filter_size = 2\n",
    "\n",
    "# Construct model \n",
    "model = Sequential()\n",
    "model.add(Conv2D(filters=16, kernel_size=filter_size,\n",
    "                 input_shape=(num_rows, num_columns, num_channels), activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=32, kernel_size=filter_size, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=64, kernel_size=filter_size, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(Conv2D(filters=128, kernel_size=filter_size, activation='relu'))\n",
    "model.add(MaxPooling2D(pool_size=2))\n",
    "model.add(Dropout(0.2))\n",
    "\n",
    "model.add(GlobalAveragePooling2D())\n",
    "\n",
    "model.add(Dense(num_labels, activation='softmax')) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compile the model\n",
    "model.compile(loss='categorical_crossentropy', metrics=['accuracy'], optimizer='adam') "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"sequential\"\n",
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d (Conv2D)              (None, 39, 861, 16)       80        \n",
      "_________________________________________________________________\n",
      "max_pooling2d (MaxPooling2D) (None, 19, 430, 16)       0         \n",
      "_________________________________________________________________\n",
      "dropout (Dropout)            (None, 19, 430, 16)       0         \n",
      "_________________________________________________________________\n",
      "conv2d_1 (Conv2D)            (None, 18, 429, 32)       2080      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 9, 214, 32)        0         \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 9, 214, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 8, 213, 64)        8256      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 4, 106, 64)        0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 4, 106, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 3, 105, 128)       32896     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 1, 52, 128)        0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 1, 52, 128)        0         \n",
      "_________________________________________________________________\n",
      "global_average_pooling2d (Gl (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense (Dense)                (None, 6)                 774       \n",
      "=================================================================\n",
      "Total params: 44,086\n",
      "Trainable params: 44,086\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "184/1 [================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================================] - 1s 7ms/sample - loss: 2.2955 - accuracy: 0.8533\n",
      "Pre-training accuracy: 85.3261%\n"
     ]
    }
   ],
   "source": [
    "# Display model architecture summary \n",
    "model.summary()\n",
    "\n",
    "# Calculate pre-training accuracy \n",
    "score = model.evaluate(x_test, y_test, verbose=1)\n",
    "accuracy = 100*score[1]\n",
    "\n",
    "print(\"Pre-training accuracy: %.4f%%\" % accuracy)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Training**\n",
    "\n",
    "Here we will train the model. If we have a trained model, we can load it instead from the next cell."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 733 samples, validate on 184 samples\n",
      "Epoch 1/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 4.6752 - accuracy: 0.6016\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.86413, saving model to mymodel2_01.h5\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 4.7966 - accuracy: 0.6317 - val_loss: 2.3647 - val_accuracy: 0.8641\n",
      "Epoch 2/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 3.8520 - accuracy: 0.8156\n",
      "Epoch 00002: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 4.1172 - accuracy: 0.8158 - val_loss: 1.9138 - val_accuracy: 0.8641\n",
      "Epoch 3/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 2.8458 - accuracy: 0.8547\n",
      "Epoch 00003: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 2.6978 - accuracy: 0.8431 - val_loss: 1.2549 - val_accuracy: 0.8533\n",
      "Epoch 4/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 1.7981 - accuracy: 0.7500\n",
      "Epoch 00004: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 1.7578 - accuracy: 0.7626 - val_loss: 0.9607 - val_accuracy: 0.8641\n",
      "Epoch 5/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 1.1436 - accuracy: 0.8547\n",
      "Epoch 00005: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 1.1222 - accuracy: 0.8527 - val_loss: 0.8839 - val_accuracy: 0.8641\n",
      "Epoch 6/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.7418 - accuracy: 0.8438\n",
      "Epoch 00006: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.7462 - accuracy: 0.8404 - val_loss: 1.0391 - val_accuracy: 0.8207\n",
      "Epoch 7/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.7029 - accuracy: 0.8531\n",
      "Epoch 00007: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.6814 - accuracy: 0.8581 - val_loss: 1.1022 - val_accuracy: 0.8533\n",
      "Epoch 8/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.6074 - accuracy: 0.8516\n",
      "Epoch 00008: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.6152 - accuracy: 0.8527 - val_loss: 1.0444 - val_accuracy: 0.8641\n",
      "Epoch 9/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.5546 - accuracy: 0.8547\n",
      "Epoch 00009: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 19s 26ms/sample - loss: 0.5814 - accuracy: 0.8486 - val_loss: 0.9323 - val_accuracy: 0.8641\n",
      "Epoch 10/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.5637 - accuracy: 0.8531\n",
      "Epoch 00010: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 17s 24ms/sample - loss: 0.5629 - accuracy: 0.8527 - val_loss: 0.9344 - val_accuracy: 0.8641\n",
      "Epoch 11/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.5193 - accuracy: 0.8484\n",
      "Epoch 00011: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.5213 - accuracy: 0.8472 - val_loss: 0.8734 - val_accuracy: 0.8641\n",
      "Epoch 12/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.5286 - accuracy: 0.8422\n",
      "Epoch 00012: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.5100 - accuracy: 0.8513 - val_loss: 0.8027 - val_accuracy: 0.8641\n",
      "Epoch 13/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.5167 - accuracy: 0.8469\n",
      "Epoch 00013: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.4998 - accuracy: 0.8513 - val_loss: 0.7569 - val_accuracy: 0.8641\n",
      "Epoch 14/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4661 - accuracy: 0.8625\n",
      "Epoch 00014: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.4765 - accuracy: 0.8540 - val_loss: 0.7103 - val_accuracy: 0.8641\n",
      "Epoch 15/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4758 - accuracy: 0.8484\n",
      "Epoch 00015: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.4623 - accuracy: 0.8527 - val_loss: 0.6986 - val_accuracy: 0.8641\n",
      "Epoch 16/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4879 - accuracy: 0.8375\n",
      "Epoch 00016: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.4685 - accuracy: 0.8472 - val_loss: 0.6701 - val_accuracy: 0.8641\n",
      "Epoch 17/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4787 - accuracy: 0.8578\n",
      "Epoch 00017: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.4726 - accuracy: 0.8622 - val_loss: 0.6236 - val_accuracy: 0.8641\n",
      "Epoch 18/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4746 - accuracy: 0.8516\n",
      "Epoch 00018: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.4646 - accuracy: 0.8554 - val_loss: 0.6382 - val_accuracy: 0.8641\n",
      "Epoch 19/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4543 - accuracy: 0.8562\n",
      "Epoch 00019: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.4601 - accuracy: 0.8540 - val_loss: 0.6130 - val_accuracy: 0.8641\n",
      "Epoch 20/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4637 - accuracy: 0.8500\n",
      "Epoch 00020: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.4418 - accuracy: 0.8622 - val_loss: 0.6274 - val_accuracy: 0.8641\n",
      "Epoch 21/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4318 - accuracy: 0.8500\n",
      "Epoch 00021: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.4256 - accuracy: 0.8554 - val_loss: 0.5971 - val_accuracy: 0.8641\n",
      "Epoch 22/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4290 - accuracy: 0.8500\n",
      "Epoch 00022: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.4147 - accuracy: 0.8581 - val_loss: 0.6110 - val_accuracy: 0.8641\n",
      "Epoch 23/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4196 - accuracy: 0.8609\n",
      "Epoch 00023: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.4378 - accuracy: 0.8581 - val_loss: 0.5929 - val_accuracy: 0.8641\n",
      "Epoch 24/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4040 - accuracy: 0.8672\n",
      "Epoch 00024: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.4166 - accuracy: 0.8649 - val_loss: 0.5955 - val_accuracy: 0.8641\n",
      "Epoch 25/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4129 - accuracy: 0.8531\n",
      "Epoch 00025: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.4143 - accuracy: 0.8540 - val_loss: 0.5888 - val_accuracy: 0.8641\n",
      "Epoch 26/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3964 - accuracy: 0.8609\n",
      "Epoch 00026: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.3960 - accuracy: 0.8568 - val_loss: 0.5790 - val_accuracy: 0.8641\n",
      "Epoch 27/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.3866 - accuracy: 0.8797\n",
      "Epoch 00027: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.3958 - accuracy: 0.8745 - val_loss: 0.5719 - val_accuracy: 0.8641\n",
      "Epoch 28/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3830 - accuracy: 0.8562\n",
      "Epoch 00028: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.3917 - accuracy: 0.8608 - val_loss: 0.5576 - val_accuracy: 0.8641\n",
      "Epoch 29/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3834 - accuracy: 0.8641\n",
      "Epoch 00029: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.3835 - accuracy: 0.8663 - val_loss: 0.5603 - val_accuracy: 0.8641\n",
      "Epoch 30/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3895 - accuracy: 0.8625\n",
      "Epoch 00030: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3809 - accuracy: 0.8663 - val_loss: 0.5525 - val_accuracy: 0.8641\n",
      "Epoch 31/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3772 - accuracy: 0.8734\n",
      "Epoch 00031: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3818 - accuracy: 0.8704 - val_loss: 0.5428 - val_accuracy: 0.8641\n",
      "Epoch 32/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4142 - accuracy: 0.8609\n",
      "Epoch 00032: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.3895 - accuracy: 0.8690 - val_loss: 0.5437 - val_accuracy: 0.8641\n",
      "Epoch 33/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3719 - accuracy: 0.8672\n",
      "Epoch 00033: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 17s 24ms/sample - loss: 0.3780 - accuracy: 0.8649 - val_loss: 0.5259 - val_accuracy: 0.8641\n",
      "Epoch 34/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.4096 - accuracy: 0.8578\n",
      "Epoch 00034: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.3852 - accuracy: 0.8677 - val_loss: 0.5287 - val_accuracy: 0.8641\n",
      "Epoch 35/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3469 - accuracy: 0.8766\n",
      "Epoch 00035: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3508 - accuracy: 0.8745 - val_loss: 0.5244 - val_accuracy: 0.8641\n",
      "Epoch 36/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3640 - accuracy: 0.8781\n",
      "Epoch 00036: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.3702 - accuracy: 0.8772 - val_loss: 0.5142 - val_accuracy: 0.8641\n",
      "Epoch 37/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3468 - accuracy: 0.8766\n",
      "Epoch 00037: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.3632 - accuracy: 0.8704 - val_loss: 0.5271 - val_accuracy: 0.8587\n",
      "Epoch 38/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3677 - accuracy: 0.8562\n",
      "Epoch 00038: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3640 - accuracy: 0.8622 - val_loss: 0.5218 - val_accuracy: 0.8641\n",
      "Epoch 39/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3477 - accuracy: 0.8891\n",
      "Epoch 00039: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3725 - accuracy: 0.8759 - val_loss: 0.5022 - val_accuracy: 0.8641\n",
      "Epoch 40/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3391 - accuracy: 0.8687\n",
      "Epoch 00040: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3487 - accuracy: 0.8690 - val_loss: 0.5088 - val_accuracy: 0.8641\n",
      "Epoch 41/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3612 - accuracy: 0.8750\n",
      "Epoch 00041: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3583 - accuracy: 0.8745 - val_loss: 0.5055 - val_accuracy: 0.8641\n",
      "Epoch 42/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3484 - accuracy: 0.8828\n",
      "Epoch 00042: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.3431 - accuracy: 0.8868 - val_loss: 0.5145 - val_accuracy: 0.8641\n",
      "Epoch 43/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3547 - accuracy: 0.8781\n",
      "Epoch 00043: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3522 - accuracy: 0.8786 - val_loss: 0.4982 - val_accuracy: 0.8587\n",
      "Epoch 44/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3446 - accuracy: 0.8750\n",
      "Epoch 00044: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.3565 - accuracy: 0.8718 - val_loss: 0.5134 - val_accuracy: 0.8641\n",
      "Epoch 45/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3585 - accuracy: 0.8734\n",
      "Epoch 00045: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3489 - accuracy: 0.8759 - val_loss: 0.4939 - val_accuracy: 0.8641\n",
      "Epoch 46/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3019 - accuracy: 0.8906\n",
      "Epoch 00046: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.3272 - accuracy: 0.8799 - val_loss: 0.4913 - val_accuracy: 0.8641\n",
      "Epoch 47/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3247 - accuracy: 0.8781\n",
      "Epoch 00047: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.3200 - accuracy: 0.8786 - val_loss: 0.4923 - val_accuracy: 0.8641\n",
      "Epoch 48/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3269 - accuracy: 0.8813\n",
      "Epoch 00048: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3238 - accuracy: 0.8840 - val_loss: 0.4953 - val_accuracy: 0.8641\n",
      "Epoch 49/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3217 - accuracy: 0.8938\n",
      "Epoch 00049: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.3345 - accuracy: 0.8881 - val_loss: 0.4873 - val_accuracy: 0.8641\n",
      "Epoch 50/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3090 - accuracy: 0.8813\n",
      "Epoch 00050: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3104 - accuracy: 0.8799 - val_loss: 0.4722 - val_accuracy: 0.8641\n",
      "Epoch 51/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2876 - accuracy: 0.8922\n",
      "Epoch 00051: val_accuracy did not improve from 0.86413\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.2990 - accuracy: 0.8895 - val_loss: 0.4698 - val_accuracy: 0.8641\n",
      "Epoch 52/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3169 - accuracy: 0.8844\n",
      "Epoch 00052: val_accuracy improved from 0.86413 to 0.86957, saving model to mymodel2_52.h5\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3014 - accuracy: 0.8909 - val_loss: 0.4636 - val_accuracy: 0.8696\n",
      "Epoch 53/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3169 - accuracy: 0.8953\n",
      "Epoch 00053: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.3148 - accuracy: 0.8936 - val_loss: 0.4644 - val_accuracy: 0.8587\n",
      "Epoch 54/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3065 - accuracy: 0.8906\n",
      "Epoch 00054: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.3162 - accuracy: 0.8827 - val_loss: 0.4594 - val_accuracy: 0.8696\n",
      "Epoch 55/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2968 - accuracy: 0.8922\n",
      "Epoch 00055: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.2878 - accuracy: 0.8963 - val_loss: 0.4643 - val_accuracy: 0.8696\n",
      "Epoch 56/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3125 - accuracy: 0.8797\n",
      "Epoch 00056: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2963 - accuracy: 0.8854 - val_loss: 0.4582 - val_accuracy: 0.8641\n",
      "Epoch 57/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.3020 - accuracy: 0.8797\n",
      "Epoch 00057: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2810 - accuracy: 0.8922 - val_loss: 0.4575 - val_accuracy: 0.8696\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 58/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.2900 - accuracy: 0.8844\n",
      "Epoch 00058: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 20s 27ms/sample - loss: 0.3010 - accuracy: 0.8854 - val_loss: 0.4354 - val_accuracy: 0.8696\n",
      "Epoch 59/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.2818 - accuracy: 0.8906\n",
      "Epoch 00059: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 19s 26ms/sample - loss: 0.2868 - accuracy: 0.8895 - val_loss: 0.4548 - val_accuracy: 0.8641\n",
      "Epoch 60/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.2871 - accuracy: 0.8875\n",
      "Epoch 00060: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.2843 - accuracy: 0.8909 - val_loss: 0.4393 - val_accuracy: 0.8696\n",
      "Epoch 61/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2645 - accuracy: 0.8984\n",
      "Epoch 00061: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2862 - accuracy: 0.8922 - val_loss: 0.4369 - val_accuracy: 0.8641\n",
      "Epoch 62/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2913 - accuracy: 0.8891\n",
      "Epoch 00062: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2823 - accuracy: 0.8922 - val_loss: 0.4644 - val_accuracy: 0.8696\n",
      "Epoch 63/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2733 - accuracy: 0.8906\n",
      "Epoch 00063: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2790 - accuracy: 0.8881 - val_loss: 0.4401 - val_accuracy: 0.8696\n",
      "Epoch 64/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2613 - accuracy: 0.9016\n",
      "Epoch 00064: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2641 - accuracy: 0.8990 - val_loss: 0.4509 - val_accuracy: 0.8696\n",
      "Epoch 65/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2470 - accuracy: 0.9125\n",
      "Epoch 00065: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.2586 - accuracy: 0.9059 - val_loss: 0.4359 - val_accuracy: 0.8696\n",
      "Epoch 66/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2550 - accuracy: 0.9047\n",
      "Epoch 00066: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2460 - accuracy: 0.9072 - val_loss: 0.4398 - val_accuracy: 0.8696\n",
      "Epoch 67/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2603 - accuracy: 0.8922\n",
      "Epoch 00067: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2479 - accuracy: 0.8977 - val_loss: 0.4256 - val_accuracy: 0.8641\n",
      "Epoch 68/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2541 - accuracy: 0.9016\n",
      "Epoch 00068: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2547 - accuracy: 0.9045 - val_loss: 0.4384 - val_accuracy: 0.8696\n",
      "Epoch 69/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2482 - accuracy: 0.9062\n",
      "Epoch 00069: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2516 - accuracy: 0.9059 - val_loss: 0.4388 - val_accuracy: 0.8641\n",
      "Epoch 70/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2342 - accuracy: 0.9109\n",
      "Epoch 00070: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2301 - accuracy: 0.9113 - val_loss: 0.4374 - val_accuracy: 0.8696\n",
      "Epoch 71/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2237 - accuracy: 0.9156\n",
      "Epoch 00071: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2396 - accuracy: 0.9100 - val_loss: 0.4585 - val_accuracy: 0.8696\n",
      "Epoch 72/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2451 - accuracy: 0.9094\n",
      "Epoch 00072: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2346 - accuracy: 0.9127 - val_loss: 0.4300 - val_accuracy: 0.8641\n",
      "Epoch 73/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2476 - accuracy: 0.8984\n",
      "Epoch 00073: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2451 - accuracy: 0.8990 - val_loss: 0.4402 - val_accuracy: 0.8587\n",
      "Epoch 74/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2218 - accuracy: 0.9109\n",
      "Epoch 00074: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2266 - accuracy: 0.9086 - val_loss: 0.4676 - val_accuracy: 0.8641\n",
      "Epoch 75/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2290 - accuracy: 0.9156\n",
      "Epoch 00075: val_accuracy did not improve from 0.86957\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2280 - accuracy: 0.9154 - val_loss: 0.4396 - val_accuracy: 0.8641\n",
      "Epoch 76/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2309 - accuracy: 0.9125\n",
      "Epoch 00076: val_accuracy improved from 0.86957 to 0.87500, saving model to mymodel2_76.h5\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2304 - accuracy: 0.9127 - val_loss: 0.4297 - val_accuracy: 0.8750\n",
      "Epoch 77/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2419 - accuracy: 0.9109\n",
      "Epoch 00077: val_accuracy did not improve from 0.87500\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2383 - accuracy: 0.9113 - val_loss: 0.5517 - val_accuracy: 0.8641\n",
      "Epoch 78/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2862 - accuracy: 0.8906\n",
      "Epoch 00078: val_accuracy did not improve from 0.87500\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2838 - accuracy: 0.8922 - val_loss: 0.4339 - val_accuracy: 0.8696\n",
      "Epoch 79/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2286 - accuracy: 0.9016\n",
      "Epoch 00079: val_accuracy did not improve from 0.87500\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2339 - accuracy: 0.9018 - val_loss: 0.4177 - val_accuracy: 0.8750\n",
      "Epoch 80/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2423 - accuracy: 0.8953\n",
      "Epoch 00080: val_accuracy did not improve from 0.87500\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2444 - accuracy: 0.8963 - val_loss: 0.4410 - val_accuracy: 0.8696\n",
      "Epoch 81/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2211 - accuracy: 0.9172\n",
      "Epoch 00081: val_accuracy improved from 0.87500 to 0.88043, saving model to mymodel2_81.h5\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2346 - accuracy: 0.9100 - val_loss: 0.4209 - val_accuracy: 0.8804\n",
      "Epoch 82/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2358 - accuracy: 0.8984\n",
      "Epoch 00082: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2340 - accuracy: 0.9018 - val_loss: 0.4272 - val_accuracy: 0.8696\n",
      "Epoch 83/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1972 - accuracy: 0.9203\n",
      "Epoch 00083: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2129 - accuracy: 0.9141 - val_loss: 0.4288 - val_accuracy: 0.8804\n",
      "Epoch 84/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2044 - accuracy: 0.9172\n",
      "Epoch 00084: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2213 - accuracy: 0.9100 - val_loss: 0.4231 - val_accuracy: 0.8696\n",
      "Epoch 85/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2314 - accuracy: 0.9109\n",
      "Epoch 00085: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2333 - accuracy: 0.9086 - val_loss: 0.4686 - val_accuracy: 0.8696\n",
      "Epoch 86/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2059 - accuracy: 0.9172\n",
      "Epoch 00086: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2146 - accuracy: 0.9141 - val_loss: 0.4163 - val_accuracy: 0.8641\n",
      "Epoch 87/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2283 - accuracy: 0.9141\n",
      "Epoch 00087: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2330 - accuracy: 0.9113 - val_loss: 0.4530 - val_accuracy: 0.8696\n",
      "Epoch 88/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2023 - accuracy: 0.9203\n",
      "Epoch 00088: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2068 - accuracy: 0.9181 - val_loss: 0.4475 - val_accuracy: 0.8696\n",
      "Epoch 89/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2110 - accuracy: 0.9156\n",
      "Epoch 00089: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2016 - accuracy: 0.9181 - val_loss: 0.4382 - val_accuracy: 0.8696\n",
      "Epoch 90/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1988 - accuracy: 0.9172\n",
      "Epoch 00090: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2027 - accuracy: 0.9168 - val_loss: 0.4428 - val_accuracy: 0.8696\n",
      "Epoch 91/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2113 - accuracy: 0.9203\n",
      "Epoch 00091: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2109 - accuracy: 0.9222 - val_loss: 0.4359 - val_accuracy: 0.8750\n",
      "Epoch 92/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1969 - accuracy: 0.9219\n",
      "Epoch 00092: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2021 - accuracy: 0.9181 - val_loss: 0.4531 - val_accuracy: 0.8696\n",
      "Epoch 93/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2148 - accuracy: 0.9266\n",
      "Epoch 00093: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2257 - accuracy: 0.9195 - val_loss: 0.4662 - val_accuracy: 0.8696\n",
      "Epoch 94/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2192 - accuracy: 0.9125\n",
      "Epoch 00094: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.2126 - accuracy: 0.9168 - val_loss: 0.4432 - val_accuracy: 0.8750\n",
      "Epoch 95/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1902 - accuracy: 0.9297\n",
      "Epoch 00095: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1940 - accuracy: 0.9304 - val_loss: 0.4692 - val_accuracy: 0.8641\n",
      "Epoch 96/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2297 - accuracy: 0.9062\n",
      "Epoch 00096: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.2192 - accuracy: 0.9100 - val_loss: 0.4343 - val_accuracy: 0.8641\n",
      "Epoch 97/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.2123 - accuracy: 0.9187\n",
      "Epoch 00097: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 19s 26ms/sample - loss: 0.2018 - accuracy: 0.9222 - val_loss: 0.4588 - val_accuracy: 0.8641\n",
      "Epoch 98/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1934 - accuracy: 0.9234\n",
      "Epoch 00098: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.1939 - accuracy: 0.9236 - val_loss: 0.4560 - val_accuracy: 0.8641\n",
      "Epoch 99/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1980 - accuracy: 0.9156\n",
      "Epoch 00099: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1885 - accuracy: 0.9209 - val_loss: 0.4699 - val_accuracy: 0.8696\n",
      "Epoch 100/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1783 - accuracy: 0.9250\n",
      "Epoch 00100: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1886 - accuracy: 0.9222 - val_loss: 0.4526 - val_accuracy: 0.8641\n",
      "Epoch 101/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2037 - accuracy: 0.9109\n",
      "Epoch 00101: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1914 - accuracy: 0.9181 - val_loss: 0.4628 - val_accuracy: 0.8641\n",
      "Epoch 102/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1709 - accuracy: 0.9391\n",
      "Epoch 00102: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1880 - accuracy: 0.9277 - val_loss: 0.4581 - val_accuracy: 0.8696\n",
      "Epoch 103/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1871 - accuracy: 0.9297\n",
      "Epoch 00103: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1842 - accuracy: 0.9304 - val_loss: 0.4952 - val_accuracy: 0.8641\n",
      "Epoch 104/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2067 - accuracy: 0.9203\n",
      "Epoch 00104: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.2089 - accuracy: 0.9181 - val_loss: 0.4476 - val_accuracy: 0.8641\n",
      "Epoch 105/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1933 - accuracy: 0.9234\n",
      "Epoch 00105: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.2043 - accuracy: 0.9195 - val_loss: 0.4667 - val_accuracy: 0.8750\n",
      "Epoch 106/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1864 - accuracy: 0.9203\n",
      "Epoch 00106: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1910 - accuracy: 0.9181 - val_loss: 0.4687 - val_accuracy: 0.8641\n",
      "Epoch 107/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1981 - accuracy: 0.9187\n",
      "Epoch 00107: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.2014 - accuracy: 0.9168 - val_loss: 0.4470 - val_accuracy: 0.8478\n",
      "Epoch 108/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.2080 - accuracy: 0.9187\n",
      "Epoch 00108: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.2044 - accuracy: 0.9195 - val_loss: 0.4813 - val_accuracy: 0.8696\n",
      "Epoch 109/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1742 - accuracy: 0.9375\n",
      "Epoch 00109: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1748 - accuracy: 0.9345 - val_loss: 0.4330 - val_accuracy: 0.8641\n",
      "Epoch 110/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1773 - accuracy: 0.9297\n",
      "Epoch 00110: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1870 - accuracy: 0.9236 - val_loss: 0.4605 - val_accuracy: 0.8587\n",
      "Epoch 111/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.1907 - accuracy: 0.9281\n",
      "Epoch 00111: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 18s 25ms/sample - loss: 0.1857 - accuracy: 0.9277 - val_loss: 0.4405 - val_accuracy: 0.8696\n",
      "Epoch 112/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.1788 - accuracy: 0.9344\n",
      "Epoch 00112: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 18s 25ms/sample - loss: 0.1795 - accuracy: 0.9304 - val_loss: 0.4470 - val_accuracy: 0.8587\n",
      "Epoch 113/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.1709 - accuracy: 0.9359\n",
      "Epoch 00113: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 19s 25ms/sample - loss: 0.1831 - accuracy: 0.9277 - val_loss: 0.4532 - val_accuracy: 0.8641\n",
      "Epoch 114/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1673 - accuracy: 0.9297\n",
      "Epoch 00114: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1726 - accuracy: 0.9291 - val_loss: 0.4651 - val_accuracy: 0.8587\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 115/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1631 - accuracy: 0.9359\n",
      "Epoch 00115: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1679 - accuracy: 0.9318 - val_loss: 0.4570 - val_accuracy: 0.8696\n",
      "Epoch 116/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1751 - accuracy: 0.9297\n",
      "Epoch 00116: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1745 - accuracy: 0.9332 - val_loss: 0.4457 - val_accuracy: 0.8641\n",
      "Epoch 117/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1654 - accuracy: 0.9344\n",
      "Epoch 00117: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1574 - accuracy: 0.9386 - val_loss: 0.4796 - val_accuracy: 0.8641\n",
      "Epoch 118/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1920 - accuracy: 0.9187\n",
      "Epoch 00118: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1800 - accuracy: 0.9236 - val_loss: 0.4652 - val_accuracy: 0.8696\n",
      "Epoch 119/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1799 - accuracy: 0.9219\n",
      "Epoch 00119: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1787 - accuracy: 0.9250 - val_loss: 0.4577 - val_accuracy: 0.8587\n",
      "Epoch 120/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1857 - accuracy: 0.9281\n",
      "Epoch 00120: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1786 - accuracy: 0.9291 - val_loss: 0.5442 - val_accuracy: 0.8587\n",
      "Epoch 121/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1780 - accuracy: 0.9281\n",
      "Epoch 00121: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1757 - accuracy: 0.9304 - val_loss: 0.4380 - val_accuracy: 0.8641\n",
      "Epoch 122/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1721 - accuracy: 0.9297\n",
      "Epoch 00122: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1750 - accuracy: 0.9277 - val_loss: 0.4791 - val_accuracy: 0.8533\n",
      "Epoch 123/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1788 - accuracy: 0.9234\n",
      "Epoch 00123: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1799 - accuracy: 0.9236 - val_loss: 0.4972 - val_accuracy: 0.8641\n",
      "Epoch 124/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1562 - accuracy: 0.9391\n",
      "Epoch 00124: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1535 - accuracy: 0.9400 - val_loss: 0.4573 - val_accuracy: 0.8641\n",
      "Epoch 125/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1596 - accuracy: 0.9406\n",
      "Epoch 00125: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1681 - accuracy: 0.9359 - val_loss: 0.4884 - val_accuracy: 0.8587\n",
      "Epoch 126/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1608 - accuracy: 0.9422\n",
      "Epoch 00126: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1590 - accuracy: 0.9413 - val_loss: 0.4802 - val_accuracy: 0.8641\n",
      "Epoch 127/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1485 - accuracy: 0.9328\n",
      "Epoch 00127: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1580 - accuracy: 0.9250 - val_loss: 0.4579 - val_accuracy: 0.8533\n",
      "Epoch 128/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1669 - accuracy: 0.9312\n",
      "Epoch 00128: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1586 - accuracy: 0.9345 - val_loss: 0.4786 - val_accuracy: 0.8696\n",
      "Epoch 129/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1582 - accuracy: 0.9344\n",
      "Epoch 00129: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1561 - accuracy: 0.9372 - val_loss: 0.4989 - val_accuracy: 0.8478\n",
      "Epoch 130/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1530 - accuracy: 0.9391\n",
      "Epoch 00130: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1586 - accuracy: 0.9372 - val_loss: 0.4929 - val_accuracy: 0.8533\n",
      "Epoch 131/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1459 - accuracy: 0.9484\n",
      "Epoch 00131: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1496 - accuracy: 0.9441 - val_loss: 0.4781 - val_accuracy: 0.8696\n",
      "Epoch 132/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1465 - accuracy: 0.9422\n",
      "Epoch 00132: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.1459 - accuracy: 0.9400 - val_loss: 0.5169 - val_accuracy: 0.8587\n",
      "Epoch 133/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1603 - accuracy: 0.9406\n",
      "Epoch 00133: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1538 - accuracy: 0.9413 - val_loss: 0.5118 - val_accuracy: 0.8641\n",
      "Epoch 134/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1535 - accuracy: 0.9328\n",
      "Epoch 00134: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1513 - accuracy: 0.9345 - val_loss: 0.4755 - val_accuracy: 0.8533\n",
      "Epoch 135/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1511 - accuracy: 0.9391\n",
      "Epoch 00135: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1550 - accuracy: 0.9332 - val_loss: 0.5139 - val_accuracy: 0.8587\n",
      "Epoch 136/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1420 - accuracy: 0.9438\n",
      "Epoch 00136: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.1461 - accuracy: 0.9441 - val_loss: 0.5094 - val_accuracy: 0.8533\n",
      "Epoch 137/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1348 - accuracy: 0.9438\n",
      "Epoch 00137: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1457 - accuracy: 0.9359 - val_loss: 0.4934 - val_accuracy: 0.8750\n",
      "Epoch 138/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1296 - accuracy: 0.9438\n",
      "Epoch 00138: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1363 - accuracy: 0.9427 - val_loss: 0.5439 - val_accuracy: 0.8587\n",
      "Epoch 139/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1448 - accuracy: 0.9469\n",
      "Epoch 00139: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1490 - accuracy: 0.9441 - val_loss: 0.4715 - val_accuracy: 0.8804\n",
      "Epoch 140/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1459 - accuracy: 0.9469\n",
      "Epoch 00140: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1480 - accuracy: 0.9441 - val_loss: 0.5862 - val_accuracy: 0.8641\n",
      "Epoch 141/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1880 - accuracy: 0.9281\n",
      "Epoch 00141: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1896 - accuracy: 0.9236 - val_loss: 0.5180 - val_accuracy: 0.8641\n",
      "Epoch 142/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1538 - accuracy: 0.9406\n",
      "Epoch 00142: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1491 - accuracy: 0.9427 - val_loss: 0.5123 - val_accuracy: 0.8533\n",
      "Epoch 143/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1376 - accuracy: 0.9516\n",
      "Epoch 00143: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1345 - accuracy: 0.9509 - val_loss: 0.5216 - val_accuracy: 0.8696\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 144/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.1635 - accuracy: 0.9203\n",
      "Epoch 00144: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 19s 25ms/sample - loss: 0.1527 - accuracy: 0.9250 - val_loss: 0.5046 - val_accuracy: 0.8641\n",
      "Epoch 145/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1378 - accuracy: 0.9469\n",
      "Epoch 00145: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1345 - accuracy: 0.9482 - val_loss: 0.5145 - val_accuracy: 0.8696\n",
      "Epoch 146/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1302 - accuracy: 0.9500\n",
      "Epoch 00146: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1370 - accuracy: 0.9468 - val_loss: 0.5236 - val_accuracy: 0.8478\n",
      "Epoch 147/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1484 - accuracy: 0.9328\n",
      "Epoch 00147: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1468 - accuracy: 0.9359 - val_loss: 0.5114 - val_accuracy: 0.8641\n",
      "Epoch 148/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1358 - accuracy: 0.9375\n",
      "Epoch 00148: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1388 - accuracy: 0.9372 - val_loss: 0.4965 - val_accuracy: 0.8641\n",
      "Epoch 149/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1239 - accuracy: 0.9594\n",
      "Epoch 00149: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1288 - accuracy: 0.9550 - val_loss: 0.5238 - val_accuracy: 0.8587\n",
      "Epoch 150/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1226 - accuracy: 0.9484\n",
      "Epoch 00150: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1252 - accuracy: 0.9454 - val_loss: 0.5233 - val_accuracy: 0.8478\n",
      "Epoch 151/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1057 - accuracy: 0.9656\n",
      "Epoch 00151: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1159 - accuracy: 0.9618 - val_loss: 0.5319 - val_accuracy: 0.8478\n",
      "Epoch 152/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1097 - accuracy: 0.9563\n",
      "Epoch 00152: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1149 - accuracy: 0.9550 - val_loss: 0.5575 - val_accuracy: 0.8587\n",
      "Epoch 153/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1462 - accuracy: 0.9406\n",
      "Epoch 00153: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1437 - accuracy: 0.9400 - val_loss: 0.5063 - val_accuracy: 0.8587\n",
      "Epoch 154/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1447 - accuracy: 0.9469\n",
      "Epoch 00154: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1517 - accuracy: 0.9427 - val_loss: 0.6021 - val_accuracy: 0.8696\n",
      "Epoch 155/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1386 - accuracy: 0.9531\n",
      "Epoch 00155: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1459 - accuracy: 0.9495 - val_loss: 0.5252 - val_accuracy: 0.8587\n",
      "Epoch 156/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1404 - accuracy: 0.9438\n",
      "Epoch 00156: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1386 - accuracy: 0.9454 - val_loss: 0.5492 - val_accuracy: 0.8641\n",
      "Epoch 157/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1036 - accuracy: 0.9688\n",
      "Epoch 00157: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1195 - accuracy: 0.9591 - val_loss: 0.5129 - val_accuracy: 0.8696\n",
      "Epoch 158/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1255 - accuracy: 0.9484\n",
      "Epoch 00158: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1240 - accuracy: 0.9495 - val_loss: 0.5442 - val_accuracy: 0.8533\n",
      "Epoch 159/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1209 - accuracy: 0.9516\n",
      "Epoch 00159: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1269 - accuracy: 0.9468 - val_loss: 0.5295 - val_accuracy: 0.8533\n",
      "Epoch 160/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1204 - accuracy: 0.9516\n",
      "Epoch 00160: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1200 - accuracy: 0.9536 - val_loss: 0.5309 - val_accuracy: 0.8696\n",
      "Epoch 161/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1217 - accuracy: 0.9500\n",
      "Epoch 00161: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1184 - accuracy: 0.9523 - val_loss: 0.5453 - val_accuracy: 0.8641\n",
      "Epoch 162/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1135 - accuracy: 0.9563\n",
      "Epoch 00162: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1181 - accuracy: 0.9550 - val_loss: 0.5240 - val_accuracy: 0.8641\n",
      "Epoch 163/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1214 - accuracy: 0.9438\n",
      "Epoch 00163: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.1337 - accuracy: 0.9359 - val_loss: 0.5694 - val_accuracy: 0.8533\n",
      "Epoch 164/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1189 - accuracy: 0.9547\n",
      "Epoch 00164: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1196 - accuracy: 0.9536 - val_loss: 0.5169 - val_accuracy: 0.8750\n",
      "Epoch 165/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1213 - accuracy: 0.9516\n",
      "Epoch 00165: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1119 - accuracy: 0.9563 - val_loss: 0.5712 - val_accuracy: 0.8587\n",
      "Epoch 166/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1099 - accuracy: 0.9563\n",
      "Epoch 00166: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.1064 - accuracy: 0.9604 - val_loss: 0.5037 - val_accuracy: 0.8750\n",
      "Epoch 167/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1253 - accuracy: 0.9531\n",
      "Epoch 00167: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1197 - accuracy: 0.9563 - val_loss: 0.5336 - val_accuracy: 0.8696\n",
      "Epoch 168/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1275 - accuracy: 0.9531\n",
      "Epoch 00168: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1222 - accuracy: 0.9536 - val_loss: 0.5393 - val_accuracy: 0.8478\n",
      "Epoch 169/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1346 - accuracy: 0.9484\n",
      "Epoch 00169: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1340 - accuracy: 0.9495 - val_loss: 0.5319 - val_accuracy: 0.8750\n",
      "Epoch 170/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1186 - accuracy: 0.9500\n",
      "Epoch 00170: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1124 - accuracy: 0.9536 - val_loss: 0.5948 - val_accuracy: 0.8533\n",
      "Epoch 171/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1277 - accuracy: 0.9531\n",
      "Epoch 00171: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1360 - accuracy: 0.9468 - val_loss: 0.5284 - val_accuracy: 0.8804\n",
      "Epoch 172/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1192 - accuracy: 0.9516\n",
      "Epoch 00172: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1150 - accuracy: 0.9523 - val_loss: 0.5720 - val_accuracy: 0.8533\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 173/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1099 - accuracy: 0.9531\n",
      "Epoch 00173: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.1031 - accuracy: 0.9577 - val_loss: 0.5291 - val_accuracy: 0.8696\n",
      "Epoch 174/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1151 - accuracy: 0.9500\n",
      "Epoch 00174: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1289 - accuracy: 0.9468 - val_loss: 0.5718 - val_accuracy: 0.8533\n",
      "Epoch 175/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1401 - accuracy: 0.9531\n",
      "Epoch 00175: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1395 - accuracy: 0.9495 - val_loss: 0.5397 - val_accuracy: 0.8641\n",
      "Epoch 176/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1188 - accuracy: 0.9516\n",
      "Epoch 00176: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1140 - accuracy: 0.9523 - val_loss: 0.4806 - val_accuracy: 0.8696\n",
      "Epoch 177/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0927 - accuracy: 0.9719\n",
      "Epoch 00177: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.0941 - accuracy: 0.9714 - val_loss: 0.5512 - val_accuracy: 0.8587\n",
      "Epoch 178/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0977 - accuracy: 0.9688\n",
      "Epoch 00178: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0966 - accuracy: 0.9673 - val_loss: 0.4928 - val_accuracy: 0.8587\n",
      "Epoch 179/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1076 - accuracy: 0.9563\n",
      "Epoch 00179: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.1131 - accuracy: 0.9509 - val_loss: 0.5513 - val_accuracy: 0.8641\n",
      "Epoch 180/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0949 - accuracy: 0.9625\n",
      "Epoch 00180: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 19s 25ms/sample - loss: 0.0972 - accuracy: 0.9632 - val_loss: 0.5457 - val_accuracy: 0.8696\n",
      "Epoch 181/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0926 - accuracy: 0.9688\n",
      "Epoch 00181: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 17s 24ms/sample - loss: 0.0956 - accuracy: 0.9659 - val_loss: 0.5542 - val_accuracy: 0.8587\n",
      "Epoch 182/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0997 - accuracy: 0.9609\n",
      "Epoch 00182: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 18s 24ms/sample - loss: 0.0984 - accuracy: 0.9618 - val_loss: 0.5781 - val_accuracy: 0.8696\n",
      "Epoch 183/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0913 - accuracy: 0.9578\n",
      "Epoch 00183: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0928 - accuracy: 0.9591 - val_loss: 0.5360 - val_accuracy: 0.8533\n",
      "Epoch 184/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0917 - accuracy: 0.9703\n",
      "Epoch 00184: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.0884 - accuracy: 0.9714 - val_loss: 0.6068 - val_accuracy: 0.8587\n",
      "Epoch 185/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1198 - accuracy: 0.9563\n",
      "Epoch 00185: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1147 - accuracy: 0.9577 - val_loss: 0.5216 - val_accuracy: 0.8696\n",
      "Epoch 186/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0921 - accuracy: 0.9672\n",
      "Epoch 00186: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1098 - accuracy: 0.9604 - val_loss: 0.5900 - val_accuracy: 0.8533\n",
      "Epoch 187/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0905 - accuracy: 0.9734\n",
      "Epoch 00187: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0893 - accuracy: 0.9727 - val_loss: 0.5601 - val_accuracy: 0.8641\n",
      "Epoch 188/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0872 - accuracy: 0.9672\n",
      "Epoch 00188: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0924 - accuracy: 0.9659 - val_loss: 0.5802 - val_accuracy: 0.8478\n",
      "Epoch 189/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0879 - accuracy: 0.9672\n",
      "Epoch 00189: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0860 - accuracy: 0.9673 - val_loss: 0.5376 - val_accuracy: 0.8641\n",
      "Epoch 190/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0824 - accuracy: 0.9609\n",
      "Epoch 00190: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.0820 - accuracy: 0.9632 - val_loss: 0.5870 - val_accuracy: 0.8533\n",
      "Epoch 191/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0819 - accuracy: 0.9672\n",
      "Epoch 00191: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.0838 - accuracy: 0.9673 - val_loss: 0.5666 - val_accuracy: 0.8641\n",
      "Epoch 192/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0812 - accuracy: 0.9703\n",
      "Epoch 00192: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0872 - accuracy: 0.9673 - val_loss: 0.5757 - val_accuracy: 0.8696\n",
      "Epoch 193/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0786 - accuracy: 0.9797\n",
      "Epoch 00193: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0852 - accuracy: 0.9754 - val_loss: 0.6082 - val_accuracy: 0.8587\n",
      "Epoch 194/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0774 - accuracy: 0.9766\n",
      "Epoch 00194: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0783 - accuracy: 0.9754 - val_loss: 0.5999 - val_accuracy: 0.8641\n",
      "Epoch 195/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0809 - accuracy: 0.9766\n",
      "Epoch 00195: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0803 - accuracy: 0.9754 - val_loss: 0.5782 - val_accuracy: 0.8750\n",
      "Epoch 196/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0785 - accuracy: 0.9719\n",
      "Epoch 00196: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0748 - accuracy: 0.9754 - val_loss: 0.5876 - val_accuracy: 0.8533\n",
      "Epoch 197/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0730 - accuracy: 0.9734\n",
      "Epoch 00197: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.0704 - accuracy: 0.9741 - val_loss: 0.5811 - val_accuracy: 0.8750\n",
      "Epoch 198/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0771 - accuracy: 0.9750\n",
      "Epoch 00198: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0776 - accuracy: 0.9754 - val_loss: 0.5914 - val_accuracy: 0.8696\n",
      "Epoch 199/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0739 - accuracy: 0.9750\n",
      "Epoch 00199: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0722 - accuracy: 0.9754 - val_loss: 0.5826 - val_accuracy: 0.8696\n",
      "Epoch 200/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0591 - accuracy: 0.9766\n",
      "Epoch 00200: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.0703 - accuracy: 0.9727 - val_loss: 0.6341 - val_accuracy: 0.8533\n",
      "Epoch 201/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0772 - accuracy: 0.9719\n",
      "Epoch 00201: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0725 - accuracy: 0.9741 - val_loss: 0.5828 - val_accuracy: 0.8696\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 202/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0692 - accuracy: 0.9719\n",
      "Epoch 00202: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0708 - accuracy: 0.9741 - val_loss: 0.6809 - val_accuracy: 0.8587\n",
      "Epoch 203/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0748 - accuracy: 0.9750\n",
      "Epoch 00203: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0733 - accuracy: 0.9754 - val_loss: 0.6014 - val_accuracy: 0.8750\n",
      "Epoch 204/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0769 - accuracy: 0.9703\n",
      "Epoch 00204: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0745 - accuracy: 0.9714 - val_loss: 0.6655 - val_accuracy: 0.8533\n",
      "Epoch 205/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0827 - accuracy: 0.9734\n",
      "Epoch 00205: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0794 - accuracy: 0.9754 - val_loss: 0.6030 - val_accuracy: 0.8641\n",
      "Epoch 206/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0699 - accuracy: 0.9781\n",
      "Epoch 00206: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0707 - accuracy: 0.9768 - val_loss: 0.6258 - val_accuracy: 0.8696\n",
      "Epoch 207/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0611 - accuracy: 0.9750\n",
      "Epoch 00207: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0618 - accuracy: 0.9768 - val_loss: 0.6288 - val_accuracy: 0.8696\n",
      "Epoch 208/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0744 - accuracy: 0.9688\n",
      "Epoch 00208: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0688 - accuracy: 0.9727 - val_loss: 0.5942 - val_accuracy: 0.8696\n",
      "Epoch 209/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0737 - accuracy: 0.9703\n",
      "Epoch 00209: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0739 - accuracy: 0.9714 - val_loss: 0.5992 - val_accuracy: 0.8750\n",
      "Epoch 210/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0798 - accuracy: 0.9703\n",
      "Epoch 00210: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.0767 - accuracy: 0.9700 - val_loss: 0.6584 - val_accuracy: 0.8587\n",
      "Epoch 211/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0793 - accuracy: 0.9703\n",
      "Epoch 00211: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0726 - accuracy: 0.9727 - val_loss: 0.6476 - val_accuracy: 0.8804\n",
      "Epoch 212/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0981 - accuracy: 0.9578\n",
      "Epoch 00212: val_accuracy did not improve from 0.88043\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.1000 - accuracy: 0.9591 - val_loss: 0.6652 - val_accuracy: 0.8804\n",
      "Epoch 213/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0706 - accuracy: 0.9672\n",
      "Epoch 00213: val_accuracy improved from 0.88043 to 0.88587, saving model to mymodel2_213.h5\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0663 - accuracy: 0.9700 - val_loss: 0.6380 - val_accuracy: 0.8859\n",
      "Epoch 214/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0691 - accuracy: 0.9719\n",
      "Epoch 00214: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 20s 27ms/sample - loss: 0.0701 - accuracy: 0.9727 - val_loss: 0.6212 - val_accuracy: 0.8696\n",
      "Epoch 215/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0714 - accuracy: 0.9734\n",
      "Epoch 00215: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.0699 - accuracy: 0.9741 - val_loss: 0.6436 - val_accuracy: 0.8804\n",
      "Epoch 216/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0709 - accuracy: 0.9812\n",
      "Epoch 00216: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0729 - accuracy: 0.9768 - val_loss: 0.6919 - val_accuracy: 0.8587\n",
      "Epoch 217/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0822 - accuracy: 0.9719\n",
      "Epoch 00217: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0749 - accuracy: 0.9754 - val_loss: 0.6549 - val_accuracy: 0.8804\n",
      "Epoch 218/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0610 - accuracy: 0.9734\n",
      "Epoch 00218: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 18s 24ms/sample - loss: 0.0588 - accuracy: 0.9754 - val_loss: 0.6626 - val_accuracy: 0.8641\n",
      "Epoch 219/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0577 - accuracy: 0.9891\n",
      "Epoch 00219: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0551 - accuracy: 0.9891 - val_loss: 0.6619 - val_accuracy: 0.8641\n",
      "Epoch 220/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0681 - accuracy: 0.9750\n",
      "Epoch 00220: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0619 - accuracy: 0.9782 - val_loss: 0.6043 - val_accuracy: 0.8750\n",
      "Epoch 221/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0575 - accuracy: 0.9750\n",
      "Epoch 00221: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 18s 24ms/sample - loss: 0.0557 - accuracy: 0.9768 - val_loss: 0.6982 - val_accuracy: 0.8641\n",
      "Epoch 222/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0534 - accuracy: 0.9812\n",
      "Epoch 00222: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 19s 25ms/sample - loss: 0.0618 - accuracy: 0.9768 - val_loss: 0.6628 - val_accuracy: 0.8859\n",
      "Epoch 223/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0804 - accuracy: 0.9719\n",
      "Epoch 00223: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 20s 27ms/sample - loss: 0.0776 - accuracy: 0.9714 - val_loss: 0.5749 - val_accuracy: 0.8804\n",
      "Epoch 224/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0646 - accuracy: 0.9750\n",
      "Epoch 00224: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 19s 26ms/sample - loss: 0.0646 - accuracy: 0.9741 - val_loss: 0.7044 - val_accuracy: 0.8859\n",
      "Epoch 225/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0737 - accuracy: 0.9672\n",
      "Epoch 00225: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 18s 24ms/sample - loss: 0.0662 - accuracy: 0.9714 - val_loss: 0.5870 - val_accuracy: 0.8750\n",
      "Epoch 226/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0886 - accuracy: 0.9688\n",
      "Epoch 00226: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 18s 25ms/sample - loss: 0.1043 - accuracy: 0.9645 - val_loss: 0.7581 - val_accuracy: 0.8859\n",
      "Epoch 227/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0977 - accuracy: 0.9672\n",
      "Epoch 00227: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 19s 26ms/sample - loss: 0.0999 - accuracy: 0.9645 - val_loss: 0.6206 - val_accuracy: 0.8750\n",
      "Epoch 228/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.1283 - accuracy: 0.9531\n",
      "Epoch 00228: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.1303 - accuracy: 0.9536 - val_loss: 0.6651 - val_accuracy: 0.8750\n",
      "Epoch 229/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0546 - accuracy: 0.9828\n",
      "Epoch 00229: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 14s 19ms/sample - loss: 0.0558 - accuracy: 0.9823 - val_loss: 0.6213 - val_accuracy: 0.8804\n",
      "Epoch 230/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0821 - accuracy: 0.9625\n",
      "Epoch 00230: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0833 - accuracy: 0.9618 - val_loss: 0.6473 - val_accuracy: 0.8859\n",
      "Epoch 231/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0651 - accuracy: 0.9766\n",
      "Epoch 00231: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0643 - accuracy: 0.9768 - val_loss: 0.6563 - val_accuracy: 0.8641\n",
      "Epoch 232/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0801 - accuracy: 0.9734\n",
      "Epoch 00232: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.0831 - accuracy: 0.9714 - val_loss: 0.6768 - val_accuracy: 0.8804\n",
      "Epoch 233/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0904 - accuracy: 0.9656\n",
      "Epoch 00233: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0885 - accuracy: 0.9673 - val_loss: 0.6461 - val_accuracy: 0.8750\n",
      "Epoch 234/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0696 - accuracy: 0.9719\n",
      "Epoch 00234: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.0690 - accuracy: 0.9741 - val_loss: 0.5911 - val_accuracy: 0.8696\n",
      "Epoch 235/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0757 - accuracy: 0.9703\n",
      "Epoch 00235: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0719 - accuracy: 0.9727 - val_loss: 0.6491 - val_accuracy: 0.8750\n",
      "Epoch 236/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0661 - accuracy: 0.9766\n",
      "Epoch 00236: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.0639 - accuracy: 0.9768 - val_loss: 0.6294 - val_accuracy: 0.8696\n",
      "Epoch 237/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0602 - accuracy: 0.9844\n",
      "Epoch 00237: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0560 - accuracy: 0.9850 - val_loss: 0.6522 - val_accuracy: 0.8859\n",
      "Epoch 238/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0493 - accuracy: 0.9812\n",
      "Epoch 00238: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0482 - accuracy: 0.9836 - val_loss: 0.6077 - val_accuracy: 0.8750\n",
      "Epoch 239/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0441 - accuracy: 0.9859\n",
      "Epoch 00239: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0488 - accuracy: 0.9850 - val_loss: 0.6480 - val_accuracy: 0.8696\n",
      "Epoch 240/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0382 - accuracy: 0.9922\n",
      "Epoch 00240: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.0402 - accuracy: 0.9905 - val_loss: 0.6931 - val_accuracy: 0.8804\n",
      "Epoch 241/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0610 - accuracy: 0.9734\n",
      "Epoch 00241: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 22ms/sample - loss: 0.0572 - accuracy: 0.9768 - val_loss: 0.6704 - val_accuracy: 0.8696\n",
      "Epoch 242/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0470 - accuracy: 0.9859\n",
      "Epoch 00242: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0516 - accuracy: 0.9823 - val_loss: 0.6921 - val_accuracy: 0.8587\n",
      "Epoch 243/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0461 - accuracy: 0.9891\n",
      "Epoch 00243: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0428 - accuracy: 0.9905 - val_loss: 0.6788 - val_accuracy: 0.8804\n",
      "Epoch 244/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0449 - accuracy: 0.9875\n",
      "Epoch 00244: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 16s 21ms/sample - loss: 0.0440 - accuracy: 0.9891 - val_loss: 0.6362 - val_accuracy: 0.8804\n",
      "Epoch 245/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0429 - accuracy: 0.9859\n",
      "Epoch 00245: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0406 - accuracy: 0.9864 - val_loss: 0.6985 - val_accuracy: 0.8641\n",
      "Epoch 246/250\n",
      "640/733 [=========================>....] - ETA: 2s - loss: 0.0461 - accuracy: 0.9844\n",
      "Epoch 00246: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 17s 23ms/sample - loss: 0.0447 - accuracy: 0.9850 - val_loss: 0.6971 - val_accuracy: 0.8750\n",
      "Epoch 247/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0415 - accuracy: 0.9859\n",
      "Epoch 00247: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0412 - accuracy: 0.9864 - val_loss: 0.6857 - val_accuracy: 0.8859\n",
      "Epoch 248/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0442 - accuracy: 0.9891\n",
      "Epoch 00248: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 21ms/sample - loss: 0.0439 - accuracy: 0.9891 - val_loss: 0.7053 - val_accuracy: 0.8750\n",
      "Epoch 249/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0325 - accuracy: 0.9937\n",
      "Epoch 00249: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 15s 20ms/sample - loss: 0.0350 - accuracy: 0.9918 - val_loss: 0.7014 - val_accuracy: 0.8750\n",
      "Epoch 250/250\n",
      "640/733 [=========================>....] - ETA: 1s - loss: 0.0386 - accuracy: 0.9859\n",
      "Epoch 00250: val_accuracy did not improve from 0.88587\n",
      "733/733 [==============================] - 14s 20ms/sample - loss: 0.0377 - accuracy: 0.9877 - val_loss: 0.7012 - val_accuracy: 0.8804\n",
      "Training completed in time:  1:03:32.594957\n"
     ]
    }
   ],
   "source": [
    "# train model\n",
    "num_epochs = 250\n",
    "num_batch_size = 128\n",
    "\n",
    "callbacks = [\n",
    "    ModelCheckpoint(\n",
    "        filepath='mymodel2_{epoch:02d}.h5',\n",
    "        # Path where to save the model\n",
    "        # The two parameters below mean that we will overwrite\n",
    "        # the current checkpoint if and only if\n",
    "        # the `val_accuracy` score has improved.\n",
    "        save_best_only=True,\n",
    "        monitor='val_accuracy',\n",
    "        verbose=1)\n",
    "]\n",
    "start = datetime.now()\n",
    "\n",
    "model.fit(x_train, y_train, batch_size=num_batch_size, epochs=num_epochs,\n",
    "          validation_data=(x_test, y_test), callbacks=callbacks, verbose=1)\n",
    "\n",
    "\n",
    "duration = datetime.now() - start\n",
    "print(\"Training completed in time: \", duration)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "**Test the model**\n",
    "\n",
    "Here we will review the accuracy of the model on both the training and test data sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Accuracy:  0.9399727\n",
      "Testing Accuracy:  0.8804348\n"
     ]
    }
   ],
   "source": [
    "# Evaluating the model on the training and testing set\n",
    "score = model.evaluate(x_train, y_train, verbose=0)\n",
    "print(\"Training Accuracy: \", score[1])\n",
    "\n",
    "score = model.evaluate(x_test, y_test, verbose=0)\n",
    "print(\"Testing Accuracy: \", score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "preds = model.predict(x_test) # label scores \n",
    "\n",
    "classpreds = np.argmax(preds, axis=1) # predicted classes \n",
    "\n",
    "y_testclass = np.argmax(y_test, axis=1) # true classes\n",
    "\n",
    "n_classes=6 # number of classes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Compute ROC curve and ROC area for each class\n",
    "fpr = dict()\n",
    "tpr = dict()\n",
    "roc_auc = dict()\n",
    "for i in range(n_classes):\n",
    "    fpr[i], tpr[i], _ = roc_curve(y_test[:, i], preds[:, i])\n",
    "    roc_auc[i] = auc(fpr[i], tpr[i])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "c_names = ['Bronchiectasis', 'Bronchiolitis', 'COPD', 'Healthy', 'Pneumonia', 'URTI']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1152x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot ROC curves\n",
    "fig, ax = plt.subplots(figsize=(16, 10))\n",
    "ax.plot([0, 1], [0, 1], 'k--')\n",
    "ax.set_xlim([0.0, 1.0])\n",
    "ax.set_ylim([0.0, 1.05])\n",
    "ax.set_xlabel('False Positive Rate')\n",
    "ax.set_ylabel('True Positive Rate')\n",
    "ax.set_title('ROC Curve for Each Class')\n",
    "for i in range(n_classes):\n",
    "    ax.plot(fpr[i], tpr[i], linewidth=3, label='ROC curve (area = %0.2f) for %s' % (roc_auc[i], c_names[i]))\n",
    "ax.legend(loc=\"best\", fontsize='x-large')\n",
    "ax.grid(alpha=.4)\n",
    "sns.despine()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "                precision    recall  f1-score   support\n",
      "\n",
      "Bronchiectasis       1.00      1.00      1.00         3\n",
      " Bronchiolitis       0.25      0.33      0.29         3\n",
      "          COPD       0.93      0.97      0.95       159\n",
      "       Healthy       0.00      0.00      0.00         7\n",
      "     Pneumonia       0.67      0.29      0.40         7\n",
      "          URTI       0.17      0.20      0.18         5\n",
      "\n",
      "      accuracy                           0.88       184\n",
      "     macro avg       0.50      0.47      0.47       184\n",
      "  weighted avg       0.86      0.88      0.87       184\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Classification Report\n",
    "print(classification_report(y_testclass, classpreds, target_names=c_names))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[  3   0   0   0   0   0]\n",
      " [  0   1   0   1   0   1]\n",
      " [  0   0 155   1   1   2]\n",
      " [  0   2   3   0   0   2]\n",
      " [  0   0   5   0   2   0]\n",
      " [  0   1   3   0   0   1]]\n"
     ]
    }
   ],
   "source": [
    "# Confusion Matrix\n",
    "print(confusion_matrix(y_testclass, classpreds))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}