2D Spectrogram model.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 30 Seconds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2019-01-26T07:40:53.664323Z",
     "start_time": "2019-01-26T07:40:50.716327Z"
    },
    "hide_input": false,
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.5/dist-packages/h5py/__init__.py:36: FutureWarning: Conversion of the second argument of issubdtype from `float` to `np.floating` is deprecated. In future, it will be treated as `np.float64 == np.dtype(float).type`.\n",
      "  from ._conv import register_converters as _register_converters\n",
      "Using TensorFlow backend.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Data: (5078, 65, 139, 1)\n",
      "Train Label:  (5078, 4)\n",
      "Vali Data:  (2032, 65, 139, 1)\n",
      "Vali Label:  (2032, 4)\n",
      "Test Data:  (3041, 65, 139, 1)\n",
      "Test Label:  (3041, 4)\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Dense, Flatten, Conv2D, MaxPooling2D, AveragePooling2D, Dropout\n",
    "from keras.layers import BatchNormalization, Activation\n",
    "from keras.utils import np_utils, multi_gpu_model\n",
    "import tensorflow as tf\n",
    "from keras.callbacks import TensorBoard\n",
    "np.set_printoptions(suppress=True)\n",
    "np.set_printoptions(threshold=np.nan)\n",
    "\n",
    "trainD = np.load(\"/home/hsiehch/30s/2D_spect/train_data.npy\")\n",
    "trainL = np.load(\"/home/hsiehch/30s/train_label.npy\")\n",
    "validationD = np.load(\"/home/hsiehch/30s/2D_spect/validate_data.npy\")\n",
    "validationL = np.load(\"/home/hsiehch/30s/validation_label.npy\")\n",
    "testD = np.load(\"/home/hsiehch/30s/2D_spect/test_data.npy\")\n",
    "testL = np.load(\"/home/hsiehch/30s/test_label.npy\")\n",
    "\n",
    "trainData = trainD.reshape((trainD.shape[0], trainD.shape[1], trainD.shape[2], 1))\n",
    "trainLabel = np_utils.to_categorical(trainL, 4)\n",
    "validationData = validationD.reshape((validationD.shape[0], validationD.shape[1], validationD.shape[2], 1))\n",
    "validationLabel = np_utils.to_categorical(validationL, 4)\n",
    "testData = testD.reshape((testD.shape[0], testD.shape[1], testD.shape[2], 1))\n",
    "testLabel = np_utils.to_categorical(testL, 4)\n",
    "\n",
    "print('Train Data:', trainData.shape)\n",
    "print('Train Label: ', trainLabel.shape)\n",
    "print('Vali Data: ', validationData.shape)\n",
    "print('Vali Label: ', validationLabel.shape)\n",
    "print('Test Data: ', testData.shape)\n",
    "print('Test Label: ', testLabel.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-19T10:07:38.271834Z",
     "start_time": "2018-07-19T10:07:38.011075Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_1 (Conv2D)            (None, 63, 137, 32)       320       \n",
      "_________________________________________________________________\n",
      "conv2d_2 (Conv2D)            (None, 61, 135, 32)       9248      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2 (None, 30, 67, 32)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_3 (Conv2D)            (None, 28, 65, 64)        18496     \n",
      "_________________________________________________________________\n",
      "conv2d_4 (Conv2D)            (None, 26, 63, 64)        36928     \n",
      "_________________________________________________________________\n",
      "dropout_1 (Dropout)          (None, 26, 63, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2 (None, 13, 31, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_5 (Conv2D)            (None, 11, 29, 64)        36928     \n",
      "_________________________________________________________________\n",
      "conv2d_6 (Conv2D)            (None, 9, 27, 64)         36928     \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 9, 27, 64)         0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2 (None, 4, 13, 64)         0         \n",
      "_________________________________________________________________\n",
      "conv2d_7 (Conv2D)            (None, 2, 11, 64)         36928     \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 2, 11, 64)         0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_4 (MaxPooling2 (None, 1, 5, 64)          0         \n",
      "_________________________________________________________________\n",
      "flatten_1 (Flatten)          (None, 320)               0         \n",
      "_________________________________________________________________\n",
      "dense_1 (Dense)              (None, 160)               51360     \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 160)               0         \n",
      "_________________________________________________________________\n",
      "dense_2 (Dense)              (None, 32)                5152      \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 4)                 132       \n",
      "=================================================================\n",
      "Total params: 232,420\n",
      "Trainable params: 232,420\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "model = Sequential() \n",
    "model.add(Conv2D(filters = 32, kernel_size = 3, data_format=\"channels_last\", input_shape = (trainData.shape[1], trainData.shape[2], 1), activation = \"relu\"))\n",
    "model.add(Conv2D(filters = 32, kernel_size = 3, activation = \"relu\"))\n",
    "model.add(MaxPooling2D(pool_size = (2,2)))\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = 3, activation = \"relu\"))\n",
    "model.add(Conv2D(filters = 64, kernel_size = 3, activation = \"relu\"))\n",
    "model.add(Dropout(0.3))\n",
    "model.add(MaxPooling2D(pool_size = (2,2)))\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = 3, activation = \"relu\"))\n",
    "model.add(Conv2D(filters = 64, kernel_size = 3, activation = \"relu\"))\n",
    "model.add(Dropout(0.3))\n",
    "model.add(MaxPooling2D(pool_size = (2,2)))\n",
    "\n",
    "model.add(Conv2D(filters = 64, kernel_size = 3, activation = \"relu\"))\n",
    "model.add(Dropout(0.3))\n",
    "model.add(MaxPooling2D(pool_size = (2,2)))\n",
    "\n",
    "model.add(Flatten())\n",
    "model.add(Dense(160, activation = \"relu\"))\n",
    "model.add(Dropout(0.5))\n",
    "model.add(Dense(32, activation = \"relu\"))\n",
    "model.add(Dense(4, activation = \"softmax\"))\n",
    "\n",
    "print(model.summary())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-19T10:15:57.494305Z",
     "start_time": "2018-07-19T10:07:41.636750Z"
    },
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 5078 samples, validate on 2032 samples\n",
      "Epoch 1/150\n",
      "5078/5078 [==============================] - 7s 1ms/step - loss: 1.0788 - acc: 0.5353 - val_loss: 1.1304 - val_acc: 0.5866\n",
      "Epoch 2/150\n",
      "5078/5078 [==============================] - 4s 694us/step - loss: 1.0119 - acc: 0.5853 - val_loss: 1.1082 - val_acc: 0.5866\n",
      "Epoch 3/150\n",
      "5078/5078 [==============================] - 3s 655us/step - loss: 1.0057 - acc: 0.5868 - val_loss: 1.0714 - val_acc: 0.5866\n",
      "Epoch 4/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 1.0017 - acc: 0.5872 - val_loss: 1.0515 - val_acc: 0.5876\n",
      "Epoch 5/150\n",
      "5078/5078 [==============================] - 3s 655us/step - loss: 0.9882 - acc: 0.5874 - val_loss: 1.0682 - val_acc: 0.5925\n",
      "Epoch 6/150\n",
      "5078/5078 [==============================] - 3s 666us/step - loss: 0.9853 - acc: 0.5857 - val_loss: 0.9864 - val_acc: 0.5896\n",
      "Epoch 7/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.9765 - acc: 0.5890 - val_loss: 0.9986 - val_acc: 0.5871\n",
      "Epoch 8/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.9707 - acc: 0.5896 - val_loss: 0.9980 - val_acc: 0.5871\n",
      "Epoch 9/150\n",
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.9706 - acc: 0.5884 - val_loss: 0.9987 - val_acc: 0.5876\n",
      "Epoch 10/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.9676 - acc: 0.5882 - val_loss: 1.0056 - val_acc: 0.5871\n",
      "Epoch 11/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.9635 - acc: 0.5902 - val_loss: 0.9964 - val_acc: 0.5866\n",
      "Epoch 12/150\n",
      "5078/5078 [==============================] - 3s 663us/step - loss: 0.9656 - acc: 0.5886 - val_loss: 0.9849 - val_acc: 0.5935\n",
      "Epoch 13/150\n",
      "5078/5078 [==============================] - 3s 670us/step - loss: 0.9615 - acc: 0.5900 - val_loss: 0.9822 - val_acc: 0.5871\n",
      "Epoch 14/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.9631 - acc: 0.5892 - val_loss: 0.9773 - val_acc: 0.5901\n",
      "Epoch 15/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.9612 - acc: 0.5918 - val_loss: 0.9797 - val_acc: 0.5871\n",
      "Epoch 16/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.9563 - acc: 0.5892 - val_loss: 0.9715 - val_acc: 0.5871\n",
      "Epoch 17/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.9568 - acc: 0.5916 - val_loss: 0.9673 - val_acc: 0.5876\n",
      "Epoch 18/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.9585 - acc: 0.5912 - val_loss: 0.9694 - val_acc: 0.5901\n",
      "Epoch 19/150\n",
      "5078/5078 [==============================] - 3s 672us/step - loss: 0.9549 - acc: 0.5931 - val_loss: 0.9638 - val_acc: 0.5861\n",
      "Epoch 20/150\n",
      "5078/5078 [==============================] - 3s 646us/step - loss: 0.9539 - acc: 0.5924 - val_loss: 0.9642 - val_acc: 0.5935\n",
      "Epoch 21/150\n",
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.9529 - acc: 0.5928 - val_loss: 0.9624 - val_acc: 0.5901\n",
      "Epoch 22/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.9488 - acc: 0.5910 - val_loss: 0.9601 - val_acc: 0.5935\n",
      "Epoch 23/150\n",
      "5078/5078 [==============================] - 3s 638us/step - loss: 0.9480 - acc: 0.5912 - val_loss: 0.9603 - val_acc: 0.5930\n",
      "Epoch 24/150\n",
      "5078/5078 [==============================] - 3s 647us/step - loss: 0.9469 - acc: 0.5914 - val_loss: 0.9591 - val_acc: 0.5861\n",
      "Epoch 25/150\n",
      "5078/5078 [==============================] - 3s 684us/step - loss: 0.9451 - acc: 0.5928 - val_loss: 0.9554 - val_acc: 0.5901\n",
      "Epoch 26/150\n",
      "5078/5078 [==============================] - 3s 656us/step - loss: 0.9434 - acc: 0.5937 - val_loss: 0.9517 - val_acc: 0.5920\n",
      "Epoch 27/150\n",
      "5078/5078 [==============================] - 3s 644us/step - loss: 0.9402 - acc: 0.5931 - val_loss: 0.9485 - val_acc: 0.5915\n",
      "Epoch 28/150\n",
      "5078/5078 [==============================] - 3s 648us/step - loss: 0.9388 - acc: 0.5953 - val_loss: 0.9547 - val_acc: 0.5881\n",
      "Epoch 29/150\n",
      "5078/5078 [==============================] - 3s 681us/step - loss: 0.9385 - acc: 0.5935 - val_loss: 0.9595 - val_acc: 0.5866\n",
      "Epoch 30/150\n",
      "5078/5078 [==============================] - 3s 632us/step - loss: 0.9381 - acc: 0.5939 - val_loss: 0.9569 - val_acc: 0.5871\n",
      "Epoch 31/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.9375 - acc: 0.5949 - val_loss: 0.9524 - val_acc: 0.5886\n",
      "Epoch 32/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.9327 - acc: 0.5953 - val_loss: 0.9420 - val_acc: 0.5930\n",
      "Epoch 33/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.9308 - acc: 0.5965 - val_loss: 0.9426 - val_acc: 0.5901\n",
      "Epoch 34/150\n",
      "5078/5078 [==============================] - 3s 671us/step - loss: 0.9253 - acc: 0.5969 - val_loss: 0.9499 - val_acc: 0.5915\n",
      "Epoch 35/150\n",
      "5078/5078 [==============================] - 3s 652us/step - loss: 0.9272 - acc: 0.5983 - val_loss: 0.9403 - val_acc: 0.5950\n",
      "Epoch 36/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.9240 - acc: 0.5994 - val_loss: 0.9383 - val_acc: 0.5910\n",
      "Epoch 37/150\n",
      "5078/5078 [==============================] - 3s 650us/step - loss: 0.9185 - acc: 0.6034 - val_loss: 0.9415 - val_acc: 0.5930\n",
      "Epoch 38/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.9186 - acc: 0.6061 - val_loss: 0.9244 - val_acc: 0.6245\n",
      "Epoch 39/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.9073 - acc: 0.6243 - val_loss: 0.9172 - val_acc: 0.6235\n",
      "Epoch 40/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.8956 - acc: 0.6286 - val_loss: 0.9128 - val_acc: 0.6255\n",
      "Epoch 41/150\n",
      "5078/5078 [==============================] - 3s 671us/step - loss: 0.8885 - acc: 0.6371 - val_loss: 0.8790 - val_acc: 0.6585\n",
      "Epoch 42/150\n",
      "5078/5078 [==============================] - 3s 665us/step - loss: 0.8478 - acc: 0.6558 - val_loss: 0.8679 - val_acc: 0.6535\n",
      "Epoch 43/150\n",
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.8341 - acc: 0.6581 - val_loss: 0.8520 - val_acc: 0.6476\n",
      "Epoch 44/150\n",
      "5078/5078 [==============================] - 3s 657us/step - loss: 0.8003 - acc: 0.6784 - val_loss: 0.7892 - val_acc: 0.6786\n",
      "Epoch 45/150\n",
      "5078/5078 [==============================] - 3s 683us/step - loss: 0.7875 - acc: 0.6761 - val_loss: 0.7979 - val_acc: 0.6708\n",
      "Epoch 46/150\n",
      "5078/5078 [==============================] - 3s 644us/step - loss: 0.7813 - acc: 0.6796 - val_loss: 0.8112 - val_acc: 0.6757\n",
      "Epoch 47/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.7712 - acc: 0.6849 - val_loss: 0.7900 - val_acc: 0.6742\n",
      "Epoch 48/150\n",
      "5078/5078 [==============================] - 3s 657us/step - loss: 0.7517 - acc: 0.6891 - val_loss: 0.7773 - val_acc: 0.6905\n",
      "Epoch 49/150\n",
      "5078/5078 [==============================] - 3s 666us/step - loss: 0.7460 - acc: 0.6938 - val_loss: 0.7860 - val_acc: 0.6683\n",
      "Epoch 50/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.7397 - acc: 0.6955 - val_loss: 0.7677 - val_acc: 0.6786\n",
      "Epoch 51/150\n",
      "5078/5078 [==============================] - 3s 631us/step - loss: 0.7218 - acc: 0.7020 - val_loss: 0.7373 - val_acc: 0.6924\n",
      "Epoch 52/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.7108 - acc: 0.7115 - val_loss: 0.7361 - val_acc: 0.7057\n",
      "Epoch 53/150\n",
      "5078/5078 [==============================] - 3s 676us/step - loss: 0.7122 - acc: 0.7024 - val_loss: 0.7186 - val_acc: 0.7116\n",
      "Epoch 54/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.6993 - acc: 0.7131 - val_loss: 0.7364 - val_acc: 0.6993\n",
      "Epoch 55/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.6954 - acc: 0.7160 - val_loss: 0.7137 - val_acc: 0.7008\n",
      "Epoch 56/150\n",
      "5078/5078 [==============================] - 3s 652us/step - loss: 0.6887 - acc: 0.7198 - val_loss: 0.6990 - val_acc: 0.7151\n",
      "Epoch 57/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.6720 - acc: 0.7221 - val_loss: 0.7267 - val_acc: 0.7032\n",
      "Epoch 58/150\n",
      "5078/5078 [==============================] - 3s 644us/step - loss: 0.6728 - acc: 0.7239 - val_loss: 0.6948 - val_acc: 0.7151\n",
      "Epoch 59/150\n",
      "5078/5078 [==============================] - 3s 657us/step - loss: 0.6619 - acc: 0.7306 - val_loss: 0.7053 - val_acc: 0.7141\n",
      "Epoch 60/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.6638 - acc: 0.7276 - val_loss: 0.7090 - val_acc: 0.7067\n",
      "Epoch 61/150\n",
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.6560 - acc: 0.7320 - val_loss: 0.6941 - val_acc: 0.7165\n",
      "Epoch 62/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.6442 - acc: 0.7341 - val_loss: 0.6721 - val_acc: 0.7239\n",
      "Epoch 63/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.6438 - acc: 0.7377 - val_loss: 0.6953 - val_acc: 0.7111\n",
      "Epoch 64/150\n",
      "5078/5078 [==============================] - 3s 633us/step - loss: 0.6312 - acc: 0.7462 - val_loss: 0.6833 - val_acc: 0.7293\n",
      "Epoch 65/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.6253 - acc: 0.7495 - val_loss: 0.6985 - val_acc: 0.7151\n",
      "Epoch 66/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.6286 - acc: 0.7393 - val_loss: 0.6772 - val_acc: 0.7170\n",
      "Epoch 67/150\n",
      "5078/5078 [==============================] - 3s 633us/step - loss: 0.6171 - acc: 0.7515 - val_loss: 0.6826 - val_acc: 0.7274\n",
      "Epoch 68/150\n",
      "5078/5078 [==============================] - 3s 652us/step - loss: 0.6094 - acc: 0.7566 - val_loss: 0.6655 - val_acc: 0.7279\n",
      "Epoch 69/150\n",
      "5078/5078 [==============================] - 3s 659us/step - loss: 0.6015 - acc: 0.7580 - val_loss: 0.6858 - val_acc: 0.7219\n",
      "Epoch 70/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.5969 - acc: 0.7607 - val_loss: 0.6935 - val_acc: 0.7279\n",
      "Epoch 71/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.5861 - acc: 0.7684 - val_loss: 0.6560 - val_acc: 0.7338\n",
      "Epoch 72/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.5937 - acc: 0.7627 - val_loss: 0.6422 - val_acc: 0.7328\n",
      "Epoch 73/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.5807 - acc: 0.7716 - val_loss: 0.6538 - val_acc: 0.7392\n",
      "Epoch 74/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.5676 - acc: 0.7822 - val_loss: 0.6621 - val_acc: 0.7333\n",
      "Epoch 75/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.5647 - acc: 0.7751 - val_loss: 0.6487 - val_acc: 0.7475\n",
      "Epoch 76/150\n",
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.5535 - acc: 0.7840 - val_loss: 0.6422 - val_acc: 0.7480\n",
      "Epoch 77/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.5581 - acc: 0.7861 - val_loss: 0.6388 - val_acc: 0.7475\n",
      "Epoch 78/150\n",
      "5078/5078 [==============================] - 3s 668us/step - loss: 0.5568 - acc: 0.7816 - val_loss: 0.6577 - val_acc: 0.7382\n",
      "Epoch 79/150\n",
      "5078/5078 [==============================] - 3s 653us/step - loss: 0.5500 - acc: 0.7848 - val_loss: 0.6542 - val_acc: 0.7446\n",
      "Epoch 80/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.5539 - acc: 0.7867 - val_loss: 0.6421 - val_acc: 0.7392\n",
      "Epoch 81/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.5435 - acc: 0.7934 - val_loss: 0.6481 - val_acc: 0.7500\n",
      "Epoch 82/150\n",
      "5078/5078 [==============================] - 3s 633us/step - loss: 0.5344 - acc: 0.7964 - val_loss: 0.6518 - val_acc: 0.7421\n",
      "Epoch 83/150\n",
      "5078/5078 [==============================] - 3s 670us/step - loss: 0.5221 - acc: 0.7972 - val_loss: 0.6525 - val_acc: 0.7530\n",
      "Epoch 84/150\n",
      "5078/5078 [==============================] - 3s 670us/step - loss: 0.5145 - acc: 0.8062 - val_loss: 0.6401 - val_acc: 0.7589\n",
      "Epoch 85/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.5187 - acc: 0.7981 - val_loss: 0.6453 - val_acc: 0.7574\n",
      "Epoch 86/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.5037 - acc: 0.8098 - val_loss: 0.6630 - val_acc: 0.7500\n",
      "Epoch 87/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.5002 - acc: 0.8135 - val_loss: 0.6261 - val_acc: 0.7667\n",
      "Epoch 88/150\n",
      "5078/5078 [==============================] - 3s 629us/step - loss: 0.4944 - acc: 0.8153 - val_loss: 0.6427 - val_acc: 0.7672\n",
      "Epoch 89/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.4950 - acc: 0.8108 - val_loss: 0.6413 - val_acc: 0.7530\n",
      "Epoch 90/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.4942 - acc: 0.8113 - val_loss: 0.6338 - val_acc: 0.7579\n",
      "Epoch 91/150\n",
      "5078/5078 [==============================] - 3s 655us/step - loss: 0.4807 - acc: 0.8208 - val_loss: 0.6350 - val_acc: 0.7603\n",
      "Epoch 92/150\n",
      "5078/5078 [==============================] - 3s 666us/step - loss: 0.4899 - acc: 0.8173 - val_loss: 0.6300 - val_acc: 0.7633\n",
      "Epoch 93/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.4901 - acc: 0.8165 - val_loss: 0.6406 - val_acc: 0.7584\n",
      "Epoch 94/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.4845 - acc: 0.8245 - val_loss: 0.6286 - val_acc: 0.7554\n",
      "Epoch 95/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.4670 - acc: 0.8271 - val_loss: 0.6109 - val_acc: 0.7677\n",
      "Epoch 96/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.4780 - acc: 0.8180 - val_loss: 0.6077 - val_acc: 0.7667\n",
      "Epoch 97/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.4648 - acc: 0.8263 - val_loss: 0.6212 - val_acc: 0.7677\n",
      "Epoch 98/150\n",
      "5078/5078 [==============================] - 3s 638us/step - loss: 0.4527 - acc: 0.8308 - val_loss: 0.6270 - val_acc: 0.7544\n",
      "Epoch 99/150\n",
      "5078/5078 [==============================] - 3s 647us/step - loss: 0.4547 - acc: 0.8314 - val_loss: 0.6015 - val_acc: 0.7662\n",
      "Epoch 100/150\n",
      "5078/5078 [==============================] - 3s 671us/step - loss: 0.4549 - acc: 0.8267 - val_loss: 0.6212 - val_acc: 0.7618\n",
      "Epoch 101/150\n",
      "5078/5078 [==============================] - 3s 654us/step - loss: 0.4538 - acc: 0.8293 - val_loss: 0.6141 - val_acc: 0.7731\n",
      "Epoch 102/150\n",
      "5078/5078 [==============================] - 3s 638us/step - loss: 0.4474 - acc: 0.8340 - val_loss: 0.6342 - val_acc: 0.7559\n",
      "Epoch 103/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.4377 - acc: 0.8387 - val_loss: 0.6378 - val_acc: 0.7534\n",
      "Epoch 104/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.4333 - acc: 0.8389 - val_loss: 0.6239 - val_acc: 0.7653\n",
      "Epoch 105/150\n",
      "5078/5078 [==============================] - 3s 633us/step - loss: 0.4303 - acc: 0.8383 - val_loss: 0.6264 - val_acc: 0.7726\n",
      "Epoch 106/150\n",
      "5078/5078 [==============================] - 3s 679us/step - loss: 0.4232 - acc: 0.8427 - val_loss: 0.6119 - val_acc: 0.7633\n",
      "Epoch 107/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.4225 - acc: 0.8407 - val_loss: 0.6400 - val_acc: 0.7584\n",
      "Epoch 108/150\n",
      "5078/5078 [==============================] - 3s 632us/step - loss: 0.4202 - acc: 0.8387 - val_loss: 0.6064 - val_acc: 0.7692\n",
      "Epoch 109/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.4151 - acc: 0.8429 - val_loss: 0.6255 - val_acc: 0.7574\n",
      "Epoch 110/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.4166 - acc: 0.8436 - val_loss: 0.6115 - val_acc: 0.7682\n",
      "Epoch 111/150\n",
      "5078/5078 [==============================] - 3s 628us/step - loss: 0.4256 - acc: 0.8434 - val_loss: 0.6191 - val_acc: 0.7682\n",
      "Epoch 112/150\n",
      "5078/5078 [==============================] - 3s 682us/step - loss: 0.4167 - acc: 0.8521 - val_loss: 0.6189 - val_acc: 0.7702\n",
      "Epoch 113/150\n",
      "5078/5078 [==============================] - 3s 649us/step - loss: 0.4106 - acc: 0.8452 - val_loss: 0.6086 - val_acc: 0.7731\n",
      "Epoch 114/150\n",
      "5078/5078 [==============================] - 3s 633us/step - loss: 0.4140 - acc: 0.8448 - val_loss: 0.6157 - val_acc: 0.7657\n",
      "Epoch 115/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.3979 - acc: 0.8549 - val_loss: 0.6024 - val_acc: 0.7712\n",
      "Epoch 116/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.4058 - acc: 0.8438 - val_loss: 0.6369 - val_acc: 0.7618\n",
      "Epoch 117/150\n",
      "5078/5078 [==============================] - 3s 633us/step - loss: 0.4026 - acc: 0.8486 - val_loss: 0.6439 - val_acc: 0.7564\n",
      "Epoch 118/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.3927 - acc: 0.8576 - val_loss: 0.6225 - val_acc: 0.7741\n",
      "Epoch 119/150\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.3871 - acc: 0.8592 - val_loss: 0.6031 - val_acc: 0.7648\n",
      "Epoch 120/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.3890 - acc: 0.8570 - val_loss: 0.6080 - val_acc: 0.7736\n",
      "Epoch 121/150\n",
      "5078/5078 [==============================] - 3s 672us/step - loss: 0.3801 - acc: 0.8562 - val_loss: 0.6247 - val_acc: 0.7608\n",
      "Epoch 122/150\n",
      "5078/5078 [==============================] - 3s 649us/step - loss: 0.3760 - acc: 0.8623 - val_loss: 0.6184 - val_acc: 0.7795\n",
      "Epoch 123/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.3765 - acc: 0.8580 - val_loss: 0.6350 - val_acc: 0.7736\n",
      "Epoch 124/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.3778 - acc: 0.8610 - val_loss: 0.6010 - val_acc: 0.7805\n",
      "Epoch 125/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.3774 - acc: 0.8627 - val_loss: 0.6501 - val_acc: 0.7687\n",
      "Epoch 126/150\n",
      "5078/5078 [==============================] - 3s 638us/step - loss: 0.3710 - acc: 0.8622 - val_loss: 0.6275 - val_acc: 0.7810\n",
      "Epoch 127/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.3606 - acc: 0.8671 - val_loss: 0.6049 - val_acc: 0.7717\n",
      "Epoch 128/150\n",
      "5078/5078 [==============================] - 3s 641us/step - loss: 0.3530 - acc: 0.8708 - val_loss: 0.6333 - val_acc: 0.7741\n",
      "Epoch 129/150\n",
      "5078/5078 [==============================] - 3s 643us/step - loss: 0.3573 - acc: 0.8629 - val_loss: 0.6392 - val_acc: 0.7731\n",
      "Epoch 130/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.3492 - acc: 0.8716 - val_loss: 0.6419 - val_acc: 0.7746\n",
      "Epoch 131/150\n",
      "5078/5078 [==============================] - 3s 650us/step - loss: 0.3599 - acc: 0.8647 - val_loss: 0.6364 - val_acc: 0.7795\n",
      "Epoch 132/150\n",
      "5078/5078 [==============================] - 3s 660us/step - loss: 0.3758 - acc: 0.8602 - val_loss: 0.6475 - val_acc: 0.7702\n",
      "Epoch 133/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.3596 - acc: 0.8623 - val_loss: 0.6965 - val_acc: 0.7692\n",
      "Epoch 134/150\n",
      "5078/5078 [==============================] - 3s 631us/step - loss: 0.3703 - acc: 0.8620 - val_loss: 0.6353 - val_acc: 0.7790\n",
      "Epoch 135/150\n",
      "5078/5078 [==============================] - 3s 642us/step - loss: 0.3571 - acc: 0.8698 - val_loss: 0.6558 - val_acc: 0.7653\n",
      "Epoch 136/150\n",
      "5078/5078 [==============================] - 3s 638us/step - loss: 0.3444 - acc: 0.8686 - val_loss: 0.6870 - val_acc: 0.7534\n",
      "Epoch 137/150\n",
      "5078/5078 [==============================] - 3s 637us/step - loss: 0.3401 - acc: 0.8748 - val_loss: 0.6663 - val_acc: 0.7731\n",
      "Epoch 138/150\n",
      "5078/5078 [==============================] - 3s 629us/step - loss: 0.3415 - acc: 0.8771 - val_loss: 0.6524 - val_acc: 0.7653\n",
      "Epoch 139/150\n",
      "5078/5078 [==============================] - 3s 636us/step - loss: 0.3499 - acc: 0.8746 - val_loss: 0.6968 - val_acc: 0.7712\n",
      "Epoch 140/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.3433 - acc: 0.8720 - val_loss: 0.6536 - val_acc: 0.7672\n",
      "Epoch 141/150\n",
      "5078/5078 [==============================] - 3s 634us/step - loss: 0.3358 - acc: 0.8755 - val_loss: 0.6761 - val_acc: 0.7756\n",
      "Epoch 142/150\n",
      "5078/5078 [==============================] - 3s 639us/step - loss: 0.3305 - acc: 0.8765 - val_loss: 0.6393 - val_acc: 0.7731\n",
      "Epoch 143/150\n",
      "5078/5078 [==============================] - 3s 629us/step - loss: 0.3317 - acc: 0.8753 - val_loss: 0.6470 - val_acc: 0.7712\n",
      "Epoch 144/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.3265 - acc: 0.8813 - val_loss: 0.6447 - val_acc: 0.7731\n",
      "Epoch 145/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.3098 - acc: 0.8844 - val_loss: 0.6852 - val_acc: 0.7726\n",
      "Epoch 146/150\n",
      "5078/5078 [==============================] - 3s 646us/step - loss: 0.3234 - acc: 0.8828 - val_loss: 0.6990 - val_acc: 0.7731\n",
      "Epoch 147/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.3025 - acc: 0.8897 - val_loss: 0.6602 - val_acc: 0.7731\n",
      "Epoch 148/150\n",
      "5078/5078 [==============================] - 3s 640us/step - loss: 0.3098 - acc: 0.8850 - val_loss: 0.6896 - val_acc: 0.7746\n",
      "Epoch 149/150\n",
      "5078/5078 [==============================] - 3s 650us/step - loss: 0.3059 - acc: 0.8883 - val_loss: 0.6803 - val_acc: 0.7657\n",
      "Epoch 150/150\n",
      "5078/5078 [==============================] - 3s 635us/step - loss: 0.3408 - acc: 0.8759 - val_loss: 0.6736 - val_acc: 0.7628\n",
      "Finish training!\n"
     ]
    }
   ],
   "source": [
    "# parallel_model = multi_gpu_model(model, gpus=8)\n",
    "model.compile(optimizer ='Adam', loss = \"categorical_crossentropy\", metrics=['accuracy'])\n",
    "batch_s = 80\n",
    "# graph = TensorBoard()\n",
    "train_history_1 = model.fit(x = trainData, \n",
    "                          y = trainLabel,\n",
    "                          epochs=150,\n",
    "                          validation_data=(validationData, validationLabel),\n",
    "                          batch_size=batch_s*1, \n",
    "                          verbose=1)\n",
    "\n",
    "print('Finish training!')\n",
    "\n",
    "import pylab as plt\n",
    "def history_display(hist, train, validation):\n",
    "    plt.plot(hist.history[train])\n",
    "    plt.plot(hist.history[validation])\n",
    "    plt.title('Train History')\n",
    "    plt.ylabel(train)\n",
    "    plt.xlabel('Epoch')\n",
    "    plt.legend(['train', 'validation'], loc='upper left')\n",
    "    plt.show\n",
    "    \n",
    "def show_plot(flag, hist):\n",
    "    if flag == 'acc':\n",
    "        history_display(hist, 'acc', 'val_acc')\n",
    "    elif flag == 'loss':\n",
    "        history_display(hist, 'loss', 'val_loss')\n",
    "    else:\n",
    "        print('Invalid!')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-19T10:16:09.561400Z",
     "start_time": "2018-07-19T10:16:09.334281Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_plot('acc', train_history_1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "ExecuteTime": {
     "end_time": "2018-07-19T10:16:15.530467Z",
     "start_time": "2018-07-19T10:16:15.322527Z"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "show_plot('loss', train_history_1)"
   ]
  }
 ],
 "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}