kinship_cnn_minor.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_10 (Conv2D)           (None, 60, 60, 16)        2416      \n",
      "_________________________________________________________________\n",
      "max_pooling2d_7 (MaxPooling2 (None, 30, 30, 16)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_11 (Conv2D)           (None, 26, 26, 64)        25664     \n",
      "_________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2 (None, 13, 13, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_12 (Conv2D)           (None, 9, 9, 128)         204928    \n",
      "_________________________________________________________________\n",
      "flatten_4 (Flatten)          (None, 10368)             0         \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 640)               6636160   \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 2)                 1282      \n",
      "=================================================================\n",
      "Total params: 6,870,450\n",
      "Trainable params: 6,870,450\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n"
     ]
    }
   ],
   "source": [
    "from keras.models import Sequential\n",
    "from keras.layers import Convolution2D\n",
    "from keras.layers import MaxPooling2D\n",
    "from keras.layers import Flatten\n",
    "from keras.layers import Dense\n",
    "from keras.optimizers import SGD\n",
    "from keras.utils import plot_model\n",
    "from keras.initializers import RandomNormal\n",
    "from matplotlib import pyplot\n",
    "\n",
    "WTInit = RandomNormal(mean=0.0, stddev=0.01, seed=5)\n",
    "model=Sequential()\n",
    "model.add(Convolution2D(16, (5, 5), input_shape=(64, 64, 6), activation=\"relu\", kernel_initializer=WTInit, bias_initializer=\"zeros\"))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Convolution2D(64, (5, 5), activation=\"relu\", kernel_initializer=WTInit, bias_initializer=\"zeros\"))\n",
    "model.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "model.add(Convolution2D(128, (5, 5), activation=\"relu\", kernel_initializer=WTInit, bias_initializer=\"zeros\"))\n",
    "\n",
    "model.add(Flatten())\n",
    "\n",
    "model.add(Dense(640, activation=\"relu\", kernel_initializer=WTInit, bias_initializer=\"zeros\"))\n",
    "model.add(Dense(2, activation=\"softmax\", kernel_initializer=WTInit, bias_initializer=\"zeros\"))\n",
    "\n",
    "sgd = SGD(lr= 0.01, momentum=0.9, decay=0.005)\n",
    "model.compile(optimizer=sgd, loss=\"categorical_crossentropy\")\n",
    "print(model.summary())\n",
    "#plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np\n",
    "import imageio\n",
    "import scipy.io\n",
    "import pandas as pd\n",
    "\n",
    "#def LoadData(DS='KinFaceW-I'):\n",
    "#mat_fd = scipy.io.loadmat(DS + '/fd_pairs.mat')\n",
    "#s = DS + '/images/'+ string[m]\n",
    "#s =  '/images/'+ string[m]\n",
    "def LoadData():\n",
    "\tcount=0\n",
    "\tall_images=[]\n",
    "\tKin = pd.DataFrame(columns=['Fold', 'Kin/Not-Kin'])\n",
    "\n",
    "\tmat_fd = scipy.io.loadmat( 'fd_pairs.mat')\n",
    "\tmat_fd = mat_fd[\"pairs\"]\n",
    "\tmat_fs = scipy.io.loadmat('fs_pairs.mat')\n",
    "\tmat_fs = mat_fs[\"pairs\"]\n",
    "\tmat_md = scipy.io.loadmat('md_pairs.mat')\n",
    "\tmat_md = mat_md[\"pairs\"]\n",
    "\tmat_ms = scipy.io.loadmat('ms_pairs.mat')\n",
    "\tmat_ms = mat_ms[\"pairs\"]\n",
    "\n",
    "\tMat = [mat_fd, mat_fs, mat_md, mat_ms]\n",
    "\tstring = ['father-dau/fd_', 'father-son/fs_', 'mother-dau/md_', 'mother-son/ms_']\n",
    "\n",
    "\tfor m in range(0, 4):\n",
    "\t\tfor j in range(0, Mat[m].shape[0]):\n",
    "\t\t\ts =  string[m]\n",
    "\t\t\taddr = s + Mat[m][j][2][0][3:6]\n",
    "\t\t\timage1 = imageio.imread(addr +'_1.jpg')\t\n",
    "\t\t\taddr = s + Mat[m][j][3][0][3:6]  \n",
    "\t\t\timage2 = imageio.imread(addr +'_2.jpg')\n",
    "\t\t\tKin.loc[count] = [Mat[m][j][0][0][0], Mat[m][j][1][0][0]]\n",
    "\t\t\tnew_image = np.concatenate((image1, image2), axis=2)\n",
    "\t\t\tall_images+=[np.array(new_image)]\n",
    "\t\t\tcount+=1\n",
    "                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                               \n",
    "\tall_images = np.array(all_images)\n",
    "\tall_images = all_images.astype('float32')\n",
    "\tall_images -= np.mean(all_images, axis=0)\n",
    "\tall_images /= np.std(all_images, axis=0)\n",
    "\tKin = np.array(Kin)\n",
    "\tData = [all_images, Kin]\n",
    "\n",
    "\trng_state = np.random.get_state()\n",
    "\tnp.random.shuffle(all_images)\n",
    "\tnp.random.set_state(rng_state)\n",
    "\tnp.random.shuffle(Kin) \n",
    "\n",
    "\tFolds = [[[], []], [[], []], [[], []], [[], []], [[], []]]\n",
    "\tfor i in range (0, all_images.shape[0]):\n",
    "\t\tfor j in range(0, 5):\n",
    "\t\t\tif(Data[1][i][0]==j+1):\n",
    "\t\t\t\tFolds[j] = np.append(Folds[j], [[np.array(all_images[i])], [np.array(Kin[i])]], axis=1)\n",
    "\n",
    "\tX=[[], [], [], [], []]\n",
    "\tY=[[], [], [], [], []] \n",
    "\tfor i in range(0, 5): \n",
    "\t\tX[i] = np.array(Folds[i][0])\n",
    "\t\tY[i] = np.array([Folds[i][1][j][1] for j in range(Folds[i].shape[1])]) \n",
    "\n",
    "\tX_Train=[[], [], [], [], []] \n",
    "\tY_Train=[[], [], [], [], []]    \n",
    "\tX_Test=[[], [], [], [], []]\n",
    "\tY_Test=[[], [], [], [], []]    \n",
    "    \n",
    "\tfor i in range(0, 5):    \n",
    "\t\tX_Train[i] = np.append(X[i%5], X[(i+1)%5], axis=0)\n",
    "\t\tX_Train[i] = np.append(X_Train[i], X[(i+2)%5], axis=0)\n",
    "\t\tX_Train[i] = np.append(X_Train[i], X[(i+3)%5], axis=0)\n",
    "\t\tY_Train[i] = np.append(Y[i%5], Y[(i+1)%5], axis=0)\n",
    "\t\tY_Train[i] = np.append(Y_Train[i], Y[(i+2)%5], axis=0)\n",
    "\t\tY_Train[i] = np.append(Y_Train[i], Y[(i+3)%5], axis=0)\n",
    "\t\tX_Test[i] = X[(i+4)%5]\n",
    "\t\tY_Test[i] = Y[(i+4)%5]\n",
    "    \n",
    "\treturn X_Train, Y_Train, X_Test, Y_Test\t"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.models import Sequential\n",
    "from keras.layers import Convolution2D\n",
    "from keras.layers import MaxPooling2D\n",
    "from keras.layers import Dropout\n",
    "from keras.layers import Activation\n",
    "from keras.layers import Flatten\n",
    "from keras.layers import Dense\n",
    "from keras.optimizers import SGD\n",
    "from keras.initializers import RandomNormal\n",
    "from keras.layers import BatchNormalization\n",
    "\n",
    "def CompileModel():\n",
    "\tmodel=Sequential()\n",
    "\tmodel.add(Convolution2D(16, (5, 5), input_shape=(64, 64, 6)))\n",
    "\tmodel.add(BatchNormalization())\n",
    "\tmodel.add(Activation('relu'))\n",
    "\tmodel.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "\tmodel.add(Convolution2D(64, (5, 5)))\n",
    "\tmodel.add(BatchNormalization())\n",
    "\tmodel.add(Activation('relu'))\n",
    "\tmodel.add(MaxPooling2D(pool_size=(2, 2)))\n",
    "\n",
    "\tmodel.add(Convolution2D(256, (5, 5)))\n",
    "\tmodel.add(BatchNormalization())\n",
    "\tmodel.add(Activation('relu'))\n",
    "\n",
    "\tmodel.add(Flatten())\n",
    "\tmodel.add(Dropout(0.5))\n",
    "    \n",
    "\tmodel.add(Dense(640))\n",
    "\tmodel.add(BatchNormalization())\n",
    "\tmodel.add(Activation('relu'))\n",
    "\n",
    "\tmodel.add(Dropout(0.5))\n",
    "\tmodel.add(Dense(2))\n",
    "\tmodel.add(Activation('softmax'))\n",
    "\n",
    "\tsgd = SGD(lr=0.0001, momentum=0.9, decay=0.005)\n",
    "\tmodel.compile(optimizer='sgd', loss=\"categorical_crossentropy\", metrics=['accuracy'])\n",
    "\tprint(model.summary())\n",
    "\treturn model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "conv2d_25 (Conv2D)           (None, 60, 60, 16)        2416      \n",
      "_________________________________________________________________\n",
      "batch_normalization_25 (Batc (None, 60, 60, 16)        64        \n",
      "_________________________________________________________________\n",
      "activation_31 (Activation)   (None, 60, 60, 16)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_17 (MaxPooling (None, 30, 30, 16)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_26 (Conv2D)           (None, 26, 26, 64)        25664     \n",
      "_________________________________________________________________\n",
      "batch_normalization_26 (Batc (None, 26, 26, 64)        256       \n",
      "_________________________________________________________________\n",
      "activation_32 (Activation)   (None, 26, 26, 64)        0         \n",
      "_________________________________________________________________\n",
      "max_pooling2d_18 (MaxPooling (None, 13, 13, 64)        0         \n",
      "_________________________________________________________________\n",
      "conv2d_27 (Conv2D)           (None, 9, 9, 256)         409856    \n",
      "_________________________________________________________________\n",
      "batch_normalization_27 (Batc (None, 9, 9, 256)         1024      \n",
      "_________________________________________________________________\n",
      "activation_33 (Activation)   (None, 9, 9, 256)         0         \n",
      "_________________________________________________________________\n",
      "flatten_9 (Flatten)          (None, 20736)             0         \n",
      "_________________________________________________________________\n",
      "dropout_13 (Dropout)         (None, 20736)             0         \n",
      "_________________________________________________________________\n",
      "dense_17 (Dense)             (None, 640)               13271680  \n",
      "_________________________________________________________________\n",
      "batch_normalization_28 (Batc (None, 640)               2560      \n",
      "_________________________________________________________________\n",
      "activation_34 (Activation)   (None, 640)               0         \n",
      "_________________________________________________________________\n",
      "dropout_14 (Dropout)         (None, 640)               0         \n",
      "_________________________________________________________________\n",
      "dense_18 (Dense)             (None, 2)                 1282      \n",
      "_________________________________________________________________\n",
      "activation_35 (Activation)   (None, 2)                 0         \n",
      "=================================================================\n",
      "Total params: 13,714,802\n",
      "Trainable params: 13,712,850\n",
      "Non-trainable params: 1,952\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 848 samples, validate on 218 samples\n",
      "Epoch 1/50\n",
      "848/848 [==============================] - 78s 92ms/step - loss: 0.9568 - acc: 0.5318 - val_loss: 0.9187 - val_acc: 0.6330\n",
      "Epoch 2/50\n",
      "848/848 [==============================] - 59s 69ms/step - loss: 0.8236 - acc: 0.5920 - val_loss: 1.4501 - val_acc: 0.5550\n",
      "Epoch 3/50\n",
      "848/848 [==============================] - 63s 74ms/step - loss: 0.7521 - acc: 0.6380 - val_loss: 0.7641 - val_acc: 0.6422\n",
      "Epoch 4/50\n",
      "848/848 [==============================] - 55s 65ms/step - loss: 0.6251 - acc: 0.6958 - val_loss: 0.6213 - val_acc: 0.7064\n",
      "Epoch 5/50\n",
      "848/848 [==============================] - 56s 66ms/step - loss: 0.5661 - acc: 0.7300 - val_loss: 0.7541 - val_acc: 0.6422\n",
      "Epoch 6/50\n",
      "848/848 [==============================] - 58s 68ms/step - loss: 0.5060 - acc: 0.7524 - val_loss: 0.8086 - val_acc: 0.6468\n",
      "Epoch 7/50\n",
      "848/848 [==============================] - 55s 64ms/step - loss: 0.4668 - acc: 0.7748 - val_loss: 0.8012 - val_acc: 0.6514\n",
      "Epoch 8/50\n",
      "848/848 [==============================] - 59s 70ms/step - loss: 0.4850 - acc: 0.7807 - val_loss: 0.5940 - val_acc: 0.7018\n",
      "Epoch 9/50\n",
      "848/848 [==============================] - 69s 82ms/step - loss: 0.4443 - acc: 0.8007 - val_loss: 0.8779 - val_acc: 0.6514\n",
      "Epoch 10/50\n",
      "848/848 [==============================] - 67s 78ms/step - loss: 0.3982 - acc: 0.8314 - val_loss: 0.5731 - val_acc: 0.7385\n",
      "Epoch 11/50\n",
      "848/848 [==============================] - 87s 103ms/step - loss: 0.3769 - acc: 0.8113 - val_loss: 0.7451 - val_acc: 0.6376\n",
      "Epoch 12/50\n",
      "848/848 [==============================] - 81s 96ms/step - loss: 0.3726 - acc: 0.8325 - val_loss: 0.6424 - val_acc: 0.6927\n",
      "Epoch 13/50\n",
      "848/848 [==============================] - 76s 90ms/step - loss: 0.3275 - acc: 0.8550 - val_loss: 0.6390 - val_acc: 0.7018\n",
      "Epoch 14/50\n",
      "848/848 [==============================] - 66s 78ms/step - loss: 0.3213 - acc: 0.8467 - val_loss: 0.6132 - val_acc: 0.7294\n",
      "Epoch 15/50\n",
      "848/848 [==============================] - 60s 70ms/step - loss: 0.2815 - acc: 0.8833 - val_loss: 0.5982 - val_acc: 0.7018\n",
      "Epoch 16/50\n",
      "848/848 [==============================] - 62s 73ms/step - loss: 0.2839 - acc: 0.8821 - val_loss: 0.5672 - val_acc: 0.7523\n",
      "Epoch 17/50\n",
      "848/848 [==============================] - 62s 73ms/step - loss: 0.2256 - acc: 0.8998 - val_loss: 0.9251 - val_acc: 0.6881\n",
      "Epoch 18/50\n",
      "848/848 [==============================] - 73s 86ms/step - loss: 0.2619 - acc: 0.8939 - val_loss: 0.6850 - val_acc: 0.7064\n",
      "Epoch 19/50\n",
      "848/848 [==============================] - 67s 79ms/step - loss: 0.2051 - acc: 0.9233 - val_loss: 0.6028 - val_acc: 0.7339\n",
      "Epoch 20/50\n",
      "848/848 [==============================] - 63s 74ms/step - loss: 0.2097 - acc: 0.9163 - val_loss: 0.7842 - val_acc: 0.7064\n",
      "Epoch 21/50\n",
      "848/848 [==============================] - 61s 72ms/step - loss: 0.1938 - acc: 0.9292 - val_loss: 0.6301 - val_acc: 0.6835\n",
      "Epoch 22/50\n",
      "848/848 [==============================] - 61s 72ms/step - loss: 0.1745 - acc: 0.9316 - val_loss: 0.6289 - val_acc: 0.7385\n",
      "Epoch 23/50\n",
      "848/848 [==============================] - 67s 80ms/step - loss: 0.1638 - acc: 0.9410 - val_loss: 0.6865 - val_acc: 0.7339\n",
      "Epoch 24/50\n",
      "848/848 [==============================] - 74s 87ms/step - loss: 0.1518 - acc: 0.9434 - val_loss: 0.6412 - val_acc: 0.7339\n",
      "Epoch 25/50\n",
      "848/848 [==============================] - 63s 74ms/step - loss: 0.1455 - acc: 0.9493 - val_loss: 0.6882 - val_acc: 0.7431\n",
      "Epoch 26/50\n",
      "848/848 [==============================] - 65s 77ms/step - loss: 0.1515 - acc: 0.9493 - val_loss: 0.6768 - val_acc: 0.7339\n",
      "Epoch 27/50\n",
      "848/848 [==============================] - 68s 80ms/step - loss: 0.1720 - acc: 0.9316 - val_loss: 0.6775 - val_acc: 0.7110\n",
      "Epoch 28/50\n",
      "848/848 [==============================] - 62s 73ms/step - loss: 0.1391 - acc: 0.9505 - val_loss: 0.7875 - val_acc: 0.7110\n",
      "Epoch 29/50\n",
      "848/848 [==============================] - 61s 72ms/step - loss: 0.1314 - acc: 0.9564 - val_loss: 0.6740 - val_acc: 0.7385\n",
      "Epoch 30/50\n",
      "848/848 [==============================] - 70s 83ms/step - loss: 0.1384 - acc: 0.9564 - val_loss: 0.6588 - val_acc: 0.7294\n",
      "Epoch 31/50\n",
      "848/848 [==============================] - 65s 77ms/step - loss: 0.1232 - acc: 0.9575 - val_loss: 0.7141 - val_acc: 0.7202\n",
      "Epoch 32/50\n",
      "848/848 [==============================] - 65s 77ms/step - loss: 0.1192 - acc: 0.9599 - val_loss: 0.6739 - val_acc: 0.7477\n",
      "Epoch 33/50\n",
      "848/848 [==============================] - 63s 75ms/step - loss: 0.1127 - acc: 0.9587 - val_loss: 0.7029 - val_acc: 0.7202\n",
      "Epoch 34/50\n",
      "848/848 [==============================] - 71s 83ms/step - loss: 0.1139 - acc: 0.9623 - val_loss: 0.6390 - val_acc: 0.7294\n",
      "Epoch 35/50\n",
      "848/848 [==============================] - 63s 74ms/step - loss: 0.1002 - acc: 0.9729 - val_loss: 0.7909 - val_acc: 0.7018\n",
      "Epoch 36/50\n",
      "848/848 [==============================] - 64s 75ms/step - loss: 0.0998 - acc: 0.9670 - val_loss: 1.3758 - val_acc: 0.6514\n",
      "Epoch 37/50\n",
      "848/848 [==============================] - 62s 73ms/step - loss: 0.1286 - acc: 0.9481 - val_loss: 0.8819 - val_acc: 0.6927\n",
      "Epoch 38/50\n",
      "848/848 [==============================] - 67s 79ms/step - loss: 0.0829 - acc: 0.9752 - val_loss: 0.7067 - val_acc: 0.7156\n",
      "Epoch 39/50\n",
      "848/848 [==============================] - 64s 75ms/step - loss: 0.0749 - acc: 0.9823 - val_loss: 0.8035 - val_acc: 0.7018\n",
      "Epoch 40/50\n",
      "848/848 [==============================] - 65s 76ms/step - loss: 0.0772 - acc: 0.9823 - val_loss: 0.7184 - val_acc: 0.7248\n",
      "Epoch 41/50\n",
      "848/848 [==============================] - 61s 72ms/step - loss: 0.0703 - acc: 0.9800 - val_loss: 0.7545 - val_acc: 0.7477\n",
      "Epoch 42/50\n",
      "848/848 [==============================] - 65s 76ms/step - loss: 0.0690 - acc: 0.9764 - val_loss: 0.6725 - val_acc: 0.7615\n",
      "Epoch 43/50\n",
      "848/848 [==============================] - 65s 77ms/step - loss: 0.0643 - acc: 0.9835 - val_loss: 0.7122 - val_acc: 0.7385\n",
      "Epoch 44/50\n",
      "848/848 [==============================] - 58s 68ms/step - loss: 0.0579 - acc: 0.9847 - val_loss: 0.8522 - val_acc: 0.7202\n",
      "Epoch 45/50\n",
      "848/848 [==============================] - 61s 72ms/step - loss: 0.0694 - acc: 0.9729 - val_loss: 0.7732 - val_acc: 0.7202\n",
      "Epoch 46/50\n",
      "848/848 [==============================] - 59s 69ms/step - loss: 0.0616 - acc: 0.9835 - val_loss: 0.7813 - val_acc: 0.7385\n",
      "Epoch 47/50\n",
      "848/848 [==============================] - 61s 72ms/step - loss: 0.0817 - acc: 0.9788 - val_loss: 0.7537 - val_acc: 0.7339\n",
      "Epoch 48/50\n",
      "848/848 [==============================] - 62s 73ms/step - loss: 0.0711 - acc: 0.9788 - val_loss: 0.6668 - val_acc: 0.7294\n",
      "Epoch 49/50\n",
      "848/848 [==============================] - 61s 71ms/step - loss: 0.0523 - acc: 0.9858 - val_loss: 0.7435 - val_acc: 0.7569\n",
      "Epoch 50/50\n",
      "848/848 [==============================] - 60s 71ms/step - loss: 0.0501 - acc: 0.9906 - val_loss: 0.7190 - val_acc: 0.7477\n",
      "218/218 [==============================] - 7s 34ms/step\n",
      "Train on 854 samples, validate on 212 samples\n",
      "Epoch 1/50\n",
      "854/854 [==============================] - 62s 72ms/step - loss: 0.2871 - acc: 0.8946 - val_loss: 0.3298 - val_acc: 0.8632\n",
      "Epoch 2/50\n",
      "854/854 [==============================] - 70s 82ms/step - loss: 0.2650 - acc: 0.9052 - val_loss: 0.0578 - val_acc: 0.9953\n",
      "Epoch 3/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.2310 - acc: 0.9110 - val_loss: 0.0556 - val_acc: 0.9811\n",
      "Epoch 4/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.1531 - acc: 0.9356 - val_loss: 0.0963 - val_acc: 0.9717\n",
      "Epoch 5/50\n",
      "854/854 [==============================] - 66s 78ms/step - loss: 0.1506 - acc: 0.9473 - val_loss: 0.0297 - val_acc: 1.0000\n",
      "Epoch 6/50\n",
      "854/854 [==============================] - 71s 83ms/step - loss: 0.1373 - acc: 0.9391 - val_loss: 0.0319 - val_acc: 1.0000\n",
      "Epoch 7/50\n",
      "854/854 [==============================] - 74s 87ms/step - loss: 0.1283 - acc: 0.9532 - val_loss: 0.0871 - val_acc: 0.9528\n",
      "Epoch 8/50\n",
      "854/854 [==============================] - 79s 93ms/step - loss: 0.1313 - acc: 0.9532 - val_loss: 0.0326 - val_acc: 0.9906\n",
      "Epoch 9/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0928 - acc: 0.9625 - val_loss: 0.0556 - val_acc: 0.9858\n",
      "Epoch 10/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0943 - acc: 0.9684 - val_loss: 0.0688 - val_acc: 0.9906\n",
      "Epoch 11/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0967 - acc: 0.9637 - val_loss: 0.0350 - val_acc: 0.9953\n",
      "Epoch 12/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0934 - acc: 0.9590 - val_loss: 0.0404 - val_acc: 0.9953\n",
      "Epoch 13/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0824 - acc: 0.9801 - val_loss: 0.0264 - val_acc: 1.0000\n",
      "Epoch 14/50\n",
      "854/854 [==============================] - 63s 73ms/step - loss: 0.0960 - acc: 0.9684 - val_loss: 0.0735 - val_acc: 0.9717\n",
      "Epoch 15/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0796 - acc: 0.9742 - val_loss: 0.0529 - val_acc: 0.9858\n",
      "Epoch 16/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0798 - acc: 0.9731 - val_loss: 0.0300 - val_acc: 1.0000\n",
      "Epoch 17/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.0815 - acc: 0.9696 - val_loss: 0.0743 - val_acc: 0.9811\n",
      "Epoch 18/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0584 - acc: 0.9848 - val_loss: 0.0376 - val_acc: 0.9953\n",
      "Epoch 19/50\n",
      "854/854 [==============================] - 56s 66ms/step - loss: 0.0675 - acc: 0.9719 - val_loss: 0.2076 - val_acc: 0.9245\n",
      "Epoch 20/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.0471 - acc: 0.9883 - val_loss: 0.0453 - val_acc: 0.9858\n",
      "Epoch 21/50\n",
      "854/854 [==============================] - 62s 73ms/step - loss: 0.0518 - acc: 0.9836 - val_loss: 0.4162 - val_acc: 0.8396\n",
      "Epoch 22/50\n",
      "854/854 [==============================] - 67s 78ms/step - loss: 0.0653 - acc: 0.9766 - val_loss: 0.2166 - val_acc: 0.8962\n",
      "Epoch 23/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0637 - acc: 0.9789 - val_loss: 0.0315 - val_acc: 0.9906\n",
      "Epoch 24/50\n",
      "854/854 [==============================] - 60s 71ms/step - loss: 0.0498 - acc: 0.9895 - val_loss: 0.0273 - val_acc: 1.0000\n",
      "Epoch 25/50\n",
      "854/854 [==============================] - 57s 67ms/step - loss: 0.0421 - acc: 0.9918 - val_loss: 0.0320 - val_acc: 0.9953\n",
      "Epoch 26/50\n",
      "854/854 [==============================] - 57s 66ms/step - loss: 0.0353 - acc: 0.9930 - val_loss: 0.0421 - val_acc: 0.9906\n",
      "Epoch 27/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0493 - acc: 0.9859 - val_loss: 0.0476 - val_acc: 0.9811\n",
      "Epoch 28/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0292 - acc: 0.9965 - val_loss: 0.0299 - val_acc: 0.9953\n",
      "Epoch 29/50\n",
      "854/854 [==============================] - 57s 66ms/step - loss: 0.0293 - acc: 0.9941 - val_loss: 0.0366 - val_acc: 0.9906\n",
      "Epoch 30/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0367 - acc: 0.9918 - val_loss: 0.0275 - val_acc: 0.9906\n",
      "Epoch 31/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0304 - acc: 0.9930 - val_loss: 0.0335 - val_acc: 0.9953\n",
      "Epoch 32/50\n",
      "854/854 [==============================] - 71s 83ms/step - loss: 0.0364 - acc: 0.9918 - val_loss: 0.0258 - val_acc: 0.9953\n",
      "Epoch 33/50\n",
      "854/854 [==============================] - 56s 66ms/step - loss: 0.0321 - acc: 0.9941 - val_loss: 0.0343 - val_acc: 0.9858\n",
      "Epoch 34/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.0289 - acc: 0.9965 - val_loss: 0.0261 - val_acc: 0.9953\n",
      "Epoch 35/50\n",
      "854/854 [==============================] - 56s 66ms/step - loss: 0.0245 - acc: 0.9941 - val_loss: 0.0333 - val_acc: 0.9858\n",
      "Epoch 36/50\n",
      "854/854 [==============================] - 62s 73ms/step - loss: 0.0302 - acc: 0.9930 - val_loss: 0.0277 - val_acc: 1.0000\n",
      "Epoch 37/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0278 - acc: 0.9930 - val_loss: 0.0554 - val_acc: 0.9764\n",
      "Epoch 38/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0194 - acc: 0.9988 - val_loss: 0.0257 - val_acc: 0.9953\n",
      "Epoch 39/50\n",
      "854/854 [==============================] - 57s 66ms/step - loss: 0.0254 - acc: 0.9930 - val_loss: 0.3965 - val_acc: 0.8868\n",
      "Epoch 40/50\n",
      "854/854 [==============================] - 61s 71ms/step - loss: 0.0484 - acc: 0.9824 - val_loss: 0.0375 - val_acc: 0.9906\n",
      "Epoch 41/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0267 - acc: 0.9906 - val_loss: 0.0253 - val_acc: 0.9953\n",
      "Epoch 42/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0260 - acc: 0.9941 - val_loss: 0.0287 - val_acc: 0.9953\n",
      "Epoch 43/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0295 - acc: 0.9930 - val_loss: 0.0292 - val_acc: 0.9953\n",
      "Epoch 44/50\n",
      "854/854 [==============================] - 76s 90ms/step - loss: 0.0205 - acc: 0.9953 - val_loss: 0.0304 - val_acc: 0.9953\n",
      "Epoch 45/50\n",
      "854/854 [==============================] - 73s 86ms/step - loss: 0.0210 - acc: 0.9953 - val_loss: 0.0385 - val_acc: 0.9906\n",
      "Epoch 46/50\n",
      "854/854 [==============================] - 60s 71ms/step - loss: 0.0177 - acc: 0.9977 - val_loss: 0.0446 - val_acc: 0.9858\n",
      "Epoch 47/50\n",
      "854/854 [==============================] - 73s 86ms/step - loss: 0.0249 - acc: 0.9930 - val_loss: 0.0306 - val_acc: 0.9906\n",
      "Epoch 48/50\n",
      "854/854 [==============================] - 62s 72ms/step - loss: 0.0203 - acc: 0.9965 - val_loss: 0.0325 - val_acc: 0.9953\n",
      "Epoch 49/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0141 - acc: 0.9977 - val_loss: 0.1830 - val_acc: 0.9434\n",
      "Epoch 50/50\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0294 - acc: 0.9941 - val_loss: 0.0367 - val_acc: 0.9906\n",
      "212/212 [==============================] - 6s 28ms/step\n",
      "Train on 854 samples, validate on 212 samples\n",
      "Epoch 1/50\n",
      "854/854 [==============================] - 59s 70ms/step - loss: 0.0616 - acc: 0.9731 - val_loss: 0.0031 - val_acc: 1.0000\n",
      "Epoch 2/50\n",
      "854/854 [==============================] - 59s 70ms/step - loss: 0.0477 - acc: 0.9836 - val_loss: 0.0078 - val_acc: 1.0000\n",
      "Epoch 3/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0379 - acc: 0.9918 - val_loss: 0.0045 - val_acc: 1.0000\n",
      "Epoch 4/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0424 - acc: 0.9918 - val_loss: 0.0021 - val_acc: 1.0000\n",
      "Epoch 5/50\n",
      "854/854 [==============================] - 63s 73ms/step - loss: 0.0312 - acc: 0.9918 - val_loss: 0.0017 - val_acc: 1.0000\n",
      "Epoch 6/50\n",
      "854/854 [==============================] - 62s 72ms/step - loss: 0.0585 - acc: 0.9754 - val_loss: 0.0047 - val_acc: 1.0000\n",
      "Epoch 7/50\n",
      "854/854 [==============================] - 57s 67ms/step - loss: 0.0333 - acc: 0.9953 - val_loss: 0.0018 - val_acc: 1.0000\n",
      "Epoch 8/50\n",
      "854/854 [==============================] - 66s 78ms/step - loss: 0.0253 - acc: 0.9941 - val_loss: 0.0016 - val_acc: 1.0000\n",
      "Epoch 9/50\n",
      "854/854 [==============================] - 65s 77ms/step - loss: 0.0154 - acc: 0.9977 - val_loss: 0.0018 - val_acc: 1.0000\n",
      "Epoch 10/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0271 - acc: 0.9941 - val_loss: 0.0081 - val_acc: 1.0000\n",
      "Epoch 11/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0306 - acc: 0.9918 - val_loss: 0.0020 - val_acc: 1.0000\n",
      "Epoch 12/50\n",
      "854/854 [==============================] - 71s 83ms/step - loss: 0.0350 - acc: 0.9883 - val_loss: 0.0319 - val_acc: 0.9906\n",
      "Epoch 13/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0401 - acc: 0.9859 - val_loss: 0.0030 - val_acc: 1.0000\n",
      "Epoch 14/50\n",
      "854/854 [==============================] - 62s 72ms/step - loss: 0.0308 - acc: 0.9918 - val_loss: 0.0021 - val_acc: 1.0000\n",
      "Epoch 15/50\n",
      "854/854 [==============================] - 67s 78ms/step - loss: 0.0247 - acc: 0.9918 - val_loss: 0.0019 - val_acc: 1.0000\n",
      "Epoch 16/50\n",
      "854/854 [==============================] - 68s 80ms/step - loss: 0.0294 - acc: 0.9930 - val_loss: 0.0020 - val_acc: 1.0000\n",
      "Epoch 17/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0235 - acc: 0.9941 - val_loss: 0.0117 - val_acc: 0.9953\n",
      "Epoch 18/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0212 - acc: 0.9953 - val_loss: 0.0015 - val_acc: 1.0000\n",
      "Epoch 19/50\n",
      "854/854 [==============================] - 72s 84ms/step - loss: 0.0242 - acc: 0.9953 - val_loss: 0.0023 - val_acc: 1.0000\n",
      "Epoch 20/50\n",
      "854/854 [==============================] - 65s 77ms/step - loss: 0.0127 - acc: 1.0000 - val_loss: 0.0017 - val_acc: 1.0000\n",
      "Epoch 21/50\n",
      "854/854 [==============================] - 71s 83ms/step - loss: 0.0186 - acc: 0.9953 - val_loss: 0.0105 - val_acc: 1.0000\n",
      "Epoch 22/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.0223 - acc: 0.9930 - val_loss: 0.0020 - val_acc: 1.0000\n",
      "Epoch 23/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0271 - acc: 0.9918 - val_loss: 0.0079 - val_acc: 1.0000\n",
      "Epoch 24/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0304 - acc: 0.9906 - val_loss: 0.0048 - val_acc: 1.0000\n",
      "Epoch 25/50\n",
      "854/854 [==============================] - 62s 73ms/step - loss: 0.0175 - acc: 0.9953 - val_loss: 0.0025 - val_acc: 1.0000\n",
      "Epoch 26/50\n",
      "854/854 [==============================] - 67s 78ms/step - loss: 0.0128 - acc: 0.9977 - val_loss: 0.0022 - val_acc: 1.0000\n",
      "Epoch 27/50\n",
      "854/854 [==============================] - 67s 79ms/step - loss: 0.0106 - acc: 0.9988 - val_loss: 0.0018 - val_acc: 1.0000\n",
      "Epoch 28/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0158 - acc: 0.9953 - val_loss: 0.2497 - val_acc: 0.9104\n",
      "Epoch 29/50\n",
      "854/854 [==============================] - 62s 73ms/step - loss: 0.0348 - acc: 0.9906 - val_loss: 0.0140 - val_acc: 0.9953\n",
      "Epoch 30/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0251 - acc: 0.9930 - val_loss: 0.0177 - val_acc: 0.9953\n",
      "Epoch 31/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0306 - acc: 0.9918 - val_loss: 0.0421 - val_acc: 0.9858\n",
      "Epoch 32/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0317 - acc: 0.9895 - val_loss: 0.0031 - val_acc: 1.0000\n",
      "Epoch 33/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0156 - acc: 0.9953 - val_loss: 0.0022 - val_acc: 1.0000\n",
      "Epoch 34/50\n",
      "854/854 [==============================] - 62s 73ms/step - loss: 0.0136 - acc: 0.9965 - val_loss: 0.0025 - val_acc: 1.0000\n",
      "Epoch 35/50\n",
      "854/854 [==============================] - 57s 66ms/step - loss: 0.0099 - acc: 1.0000 - val_loss: 0.0022 - val_acc: 1.0000\n",
      "Epoch 36/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0121 - acc: 0.9977 - val_loss: 0.0023 - val_acc: 1.0000\n",
      "Epoch 37/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0119 - acc: 1.0000 - val_loss: 0.0024 - val_acc: 1.0000\n",
      "Epoch 38/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0106 - acc: 0.9977 - val_loss: 0.0046 - val_acc: 1.0000\n",
      "Epoch 39/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0154 - acc: 0.9977 - val_loss: 0.0024 - val_acc: 1.0000\n",
      "Epoch 40/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0165 - acc: 0.9977 - val_loss: 0.0026 - val_acc: 1.0000\n",
      "Epoch 41/50\n",
      "854/854 [==============================] - 61s 71ms/step - loss: 0.0062 - acc: 0.9988 - val_loss: 0.0018 - val_acc: 1.0000\n",
      "Epoch 42/50\n",
      "854/854 [==============================] - 57s 67ms/step - loss: 0.0111 - acc: 0.9988 - val_loss: 0.0021 - val_acc: 1.0000\n",
      "Epoch 43/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0108 - acc: 1.0000 - val_loss: 0.0027 - val_acc: 1.0000\n",
      "Epoch 44/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0101 - acc: 0.9988 - val_loss: 0.0123 - val_acc: 0.9953\n",
      "Epoch 45/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0183 - acc: 0.9941 - val_loss: 0.0061 - val_acc: 1.0000\n",
      "Epoch 46/50\n",
      "854/854 [==============================] - 54s 64ms/step - loss: 0.0147 - acc: 0.9965 - val_loss: 0.0030 - val_acc: 1.0000\n",
      "Epoch 47/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0085 - acc: 0.9988 - val_loss: 0.0170 - val_acc: 0.9953\n",
      "Epoch 48/50\n",
      "854/854 [==============================] - 57s 67ms/step - loss: 0.0113 - acc: 0.9965 - val_loss: 0.0039 - val_acc: 1.0000\n",
      "Epoch 49/50\n",
      "854/854 [==============================] - 56s 66ms/step - loss: 0.0094 - acc: 0.9988 - val_loss: 0.0021 - val_acc: 1.0000\n",
      "Epoch 50/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0096 - acc: 0.9988 - val_loss: 0.0019 - val_acc: 1.0000\n",
      "212/212 [==============================] - 5s 26ms/step\n",
      "Train on 854 samples, validate on 212 samples\n",
      "Epoch 1/50\n",
      "854/854 [==============================] - 55s 65ms/step - loss: 0.0313 - acc: 0.9918 - val_loss: 5.7304e-04 - val_acc: 1.0000\n",
      "Epoch 2/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0176 - acc: 0.9965 - val_loss: 0.0346 - val_acc: 0.9953\n",
      "Epoch 3/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0188 - acc: 0.9965 - val_loss: 7.9695e-04 - val_acc: 1.0000\n",
      "Epoch 4/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0145 - acc: 0.9988 - val_loss: 5.1984e-04 - val_acc: 1.0000\n",
      "Epoch 5/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0173 - acc: 0.9953 - val_loss: 4.1920e-04 - val_acc: 1.0000\n",
      "Epoch 6/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0147 - acc: 0.9977 - val_loss: 3.4714e-04 - val_acc: 1.0000\n",
      "Epoch 7/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0153 - acc: 0.9953 - val_loss: 3.2621e-04 - val_acc: 1.0000\n",
      "Epoch 8/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0068 - acc: 1.0000 - val_loss: 3.0294e-04 - val_acc: 1.0000\n",
      "Epoch 9/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0089 - acc: 1.0000 - val_loss: 8.3558e-04 - val_acc: 1.0000\n",
      "Epoch 10/50\n",
      "854/854 [==============================] - 54s 64ms/step - loss: 0.0117 - acc: 0.9988 - val_loss: 2.9902e-04 - val_acc: 1.0000\n",
      "Epoch 11/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0139 - acc: 0.9965 - val_loss: 3.9123e-04 - val_acc: 1.0000\n",
      "Epoch 12/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0213 - acc: 0.9953 - val_loss: 0.5688 - val_acc: 0.8821\n",
      "Epoch 13/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0457 - acc: 0.9871 - val_loss: 8.7641e-04 - val_acc: 1.0000\n",
      "Epoch 14/50\n",
      "854/854 [==============================] - 54s 64ms/step - loss: 0.0255 - acc: 0.9930 - val_loss: 8.4827e-04 - val_acc: 1.0000\n",
      "Epoch 15/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0104 - acc: 0.9988 - val_loss: 6.8953e-04 - val_acc: 1.0000\n",
      "Epoch 16/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0195 - acc: 0.9941 - val_loss: 0.3131 - val_acc: 0.8632\n",
      "Epoch 17/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0362 - acc: 0.9859 - val_loss: 0.0024 - val_acc: 1.0000\n",
      "Epoch 18/50\n",
      "854/854 [==============================] - 54s 63ms/step - loss: 0.0270 - acc: 0.9930 - val_loss: 0.1382 - val_acc: 0.9340\n",
      "Epoch 19/50\n",
      "854/854 [==============================] - 62s 72ms/step - loss: 0.0274 - acc: 0.9941 - val_loss: 8.5913e-04 - val_acc: 1.0000\n",
      "Epoch 20/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.0138 - acc: 0.9988 - val_loss: 0.0010 - val_acc: 1.0000\n",
      "Epoch 21/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0140 - acc: 0.9965 - val_loss: 5.3264e-04 - val_acc: 1.0000\n",
      "Epoch 22/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0139 - acc: 0.9977 - val_loss: 0.0018 - val_acc: 1.0000\n",
      "Epoch 23/50\n",
      "854/854 [==============================] - 63s 74ms/step - loss: 0.0142 - acc: 0.9977 - val_loss: 0.0026 - val_acc: 1.0000\n",
      "Epoch 24/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0214 - acc: 0.9930 - val_loss: 9.3857e-04 - val_acc: 1.0000\n",
      "Epoch 25/50\n",
      "854/854 [==============================] - 55s 64ms/step - loss: 0.0109 - acc: 0.9988 - val_loss: 5.0814e-04 - val_acc: 1.0000\n",
      "Epoch 26/50\n",
      "854/854 [==============================] - 57s 67ms/step - loss: 0.0091 - acc: 0.9988 - val_loss: 4.7313e-04 - val_acc: 1.0000\n",
      "Epoch 27/50\n",
      "854/854 [==============================] - 77s 91ms/step - loss: 0.0158 - acc: 0.9930 - val_loss: 0.0037 - val_acc: 1.0000\n",
      "Epoch 28/50\n",
      "854/854 [==============================] - 75s 88ms/step - loss: 0.0114 - acc: 0.9965 - val_loss: 6.4846e-04 - val_acc: 1.0000\n",
      "Epoch 29/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0124 - acc: 0.9953 - val_loss: 5.2094e-04 - val_acc: 1.0000\n",
      "Epoch 30/50\n",
      "854/854 [==============================] - 67s 79ms/step - loss: 0.0152 - acc: 0.9953 - val_loss: 4.3754e-04 - val_acc: 1.0000\n",
      "Epoch 31/50\n",
      "854/854 [==============================] - 68s 80ms/step - loss: 0.0099 - acc: 0.9988 - val_loss: 5.4533e-04 - val_acc: 1.0000\n",
      "Epoch 32/50\n",
      "854/854 [==============================] - 74s 86ms/step - loss: 0.0119 - acc: 0.9953 - val_loss: 0.0011 - val_acc: 1.0000\n",
      "Epoch 33/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0172 - acc: 0.9953 - val_loss: 0.0013 - val_acc: 1.0000\n",
      "Epoch 34/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0111 - acc: 0.9965 - val_loss: 6.2924e-04 - val_acc: 1.0000\n",
      "Epoch 35/50\n",
      "854/854 [==============================] - 66s 77ms/step - loss: 0.0166 - acc: 0.9941 - val_loss: 0.2060 - val_acc: 0.9151\n",
      "Epoch 36/50\n",
      "854/854 [==============================] - 66s 77ms/step - loss: 0.0441 - acc: 0.9895 - val_loss: 0.0023 - val_acc: 1.0000\n",
      "Epoch 37/50\n",
      "854/854 [==============================] - 66s 77ms/step - loss: 0.0119 - acc: 0.9977 - val_loss: 7.3988e-04 - val_acc: 1.0000\n",
      "Epoch 38/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0141 - acc: 0.9977 - val_loss: 7.4591e-04 - val_acc: 1.0000\n",
      "Epoch 39/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0126 - acc: 0.9965 - val_loss: 0.0023 - val_acc: 1.0000\n",
      "Epoch 40/50\n",
      "854/854 [==============================] - 67s 78ms/step - loss: 0.0080 - acc: 0.9988 - val_loss: 5.0172e-04 - val_acc: 1.0000\n",
      "Epoch 41/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0130 - acc: 0.9977 - val_loss: 0.0011 - val_acc: 1.0000\n",
      "Epoch 42/50\n",
      "854/854 [==============================] - 67s 78ms/step - loss: 0.0184 - acc: 0.9941 - val_loss: 7.1017e-04 - val_acc: 1.0000\n",
      "Epoch 43/50\n",
      "854/854 [==============================] - 66s 77ms/step - loss: 0.0225 - acc: 0.9930 - val_loss: 6.8105e-04 - val_acc: 1.0000\n",
      "Epoch 44/50\n",
      "854/854 [==============================] - 67s 79ms/step - loss: 0.0114 - acc: 0.9965 - val_loss: 5.7330e-04 - val_acc: 1.0000\n",
      "Epoch 45/50\n",
      "854/854 [==============================] - 64s 74ms/step - loss: 0.0093 - acc: 0.9965 - val_loss: 5.2201e-04 - val_acc: 1.0000\n",
      "Epoch 46/50\n",
      "854/854 [==============================] - 73s 85ms/step - loss: 0.0125 - acc: 0.9965 - val_loss: 5.3921e-04 - val_acc: 1.0000\n",
      "Epoch 47/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0067 - acc: 0.9977 - val_loss: 0.0147 - val_acc: 0.9953\n",
      "Epoch 48/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0162 - acc: 0.9941 - val_loss: 0.0016 - val_acc: 1.0000\n",
      "Epoch 49/50\n",
      "854/854 [==============================] - 65s 76ms/step - loss: 0.0066 - acc: 1.0000 - val_loss: 8.8052e-04 - val_acc: 1.0000\n",
      "Epoch 50/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0084 - acc: 0.9977 - val_loss: 5.1574e-04 - val_acc: 1.0000\n",
      "212/212 [==============================] - 8s 36ms/step\n",
      "Train on 854 samples, validate on 212 samples\n",
      "Epoch 1/50\n",
      "854/854 [==============================] - 67s 79ms/step - loss: 0.0130 - acc: 0.9965 - val_loss: 2.3529e-04 - val_acc: 1.0000\n",
      "Epoch 2/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0064 - acc: 1.0000 - val_loss: 1.7680e-04 - val_acc: 1.0000\n",
      "Epoch 3/50\n",
      "854/854 [==============================] - 66s 77ms/step - loss: 0.0096 - acc: 0.9988 - val_loss: 3.8488e-04 - val_acc: 1.0000\n",
      "Epoch 4/50\n",
      "854/854 [==============================] - 69s 81ms/step - loss: 0.0181 - acc: 0.9918 - val_loss: 5.3204e-04 - val_acc: 1.0000\n",
      "Epoch 5/50\n",
      "854/854 [==============================] - 72s 84ms/step - loss: 0.0120 - acc: 0.9941 - val_loss: 1.5580e-04 - val_acc: 1.0000\n",
      "Epoch 6/50\n",
      "854/854 [==============================] - 64s 75ms/step - loss: 0.0100 - acc: 0.9965 - val_loss: 1.4140e-04 - val_acc: 1.0000\n",
      "Epoch 7/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0048 - acc: 1.0000 - val_loss: 2.0646e-04 - val_acc: 1.0000\n",
      "Epoch 8/50\n",
      "854/854 [==============================] - 90s 105ms/step - loss: 0.0194 - acc: 0.9941 - val_loss: 1.6040e-04 - val_acc: 1.0000\n",
      "Epoch 9/50\n",
      "854/854 [==============================] - 74s 87ms/step - loss: 0.0089 - acc: 0.9988 - val_loss: 1.8941e-04 - val_acc: 1.0000\n",
      "Epoch 10/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0070 - acc: 1.0000 - val_loss: 2.4449e-04 - val_acc: 1.0000\n",
      "Epoch 11/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0073 - acc: 1.0000 - val_loss: 2.2337e-04 - val_acc: 1.0000\n",
      "Epoch 12/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0057 - acc: 1.0000 - val_loss: 1.4659e-04 - val_acc: 1.0000\n",
      "Epoch 13/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0062 - acc: 1.0000 - val_loss: 1.4169e-04 - val_acc: 1.0000\n",
      "Epoch 14/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0062 - acc: 0.9988 - val_loss: 5.5172e-04 - val_acc: 1.0000\n",
      "Epoch 15/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0043 - acc: 1.0000 - val_loss: 2.3976e-04 - val_acc: 1.0000\n",
      "Epoch 16/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0104 - acc: 0.9965 - val_loss: 2.0808e-04 - val_acc: 1.0000\n",
      "Epoch 17/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0056 - acc: 0.9988 - val_loss: 2.1069e-04 - val_acc: 1.0000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 18/50\n",
      "854/854 [==============================] - 61s 72ms/step - loss: 0.0051 - acc: 0.9988 - val_loss: 2.9237e-04 - val_acc: 1.0000\n",
      "Epoch 19/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0051 - acc: 1.0000 - val_loss: 2.3048e-04 - val_acc: 1.0000\n",
      "Epoch 20/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0045 - acc: 1.0000 - val_loss: 1.7905e-04 - val_acc: 1.0000\n",
      "Epoch 21/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0049 - acc: 0.9988 - val_loss: 1.5425e-04 - val_acc: 1.0000\n",
      "Epoch 22/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0069 - acc: 0.9988 - val_loss: 2.2884e-04 - val_acc: 1.0000\n",
      "Epoch 23/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0080 - acc: 0.9988 - val_loss: 1.5528e-04 - val_acc: 1.0000\n",
      "Epoch 24/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0048 - acc: 0.9988 - val_loss: 1.7890e-04 - val_acc: 1.0000\n",
      "Epoch 25/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0053 - acc: 0.9988 - val_loss: 1.5633e-04 - val_acc: 1.0000\n",
      "Epoch 26/50\n",
      "854/854 [==============================] - 57s 67ms/step - loss: 0.0067 - acc: 0.9988 - val_loss: 1.4288e-04 - val_acc: 1.0000\n",
      "Epoch 27/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0073 - acc: 0.9965 - val_loss: 1.4953e-04 - val_acc: 1.0000\n",
      "Epoch 28/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0068 - acc: 0.9988 - val_loss: 2.0033e-04 - val_acc: 1.0000\n",
      "Epoch 29/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0049 - acc: 1.0000 - val_loss: 1.4105e-04 - val_acc: 1.0000\n",
      "Epoch 30/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0069 - acc: 1.0000 - val_loss: 4.7981e-04 - val_acc: 1.0000\n",
      "Epoch 31/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0051 - acc: 0.9988 - val_loss: 4.2077e-04 - val_acc: 1.0000\n",
      "Epoch 32/50\n",
      "854/854 [==============================] - 62s 72ms/step - loss: 0.0104 - acc: 0.9977 - val_loss: 0.0712 - val_acc: 0.9717\n",
      "Epoch 33/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0227 - acc: 0.9895 - val_loss: 3.3309e-04 - val_acc: 1.0000\n",
      "Epoch 34/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0056 - acc: 1.0000 - val_loss: 0.0044 - val_acc: 1.0000\n",
      "Epoch 35/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0092 - acc: 0.9965 - val_loss: 6.2932e-04 - val_acc: 1.0000\n",
      "Epoch 36/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0090 - acc: 0.9977 - val_loss: 7.4412e-04 - val_acc: 1.0000\n",
      "Epoch 37/50\n",
      "854/854 [==============================] - 59s 70ms/step - loss: 0.0090 - acc: 0.9988 - val_loss: 3.0959e-04 - val_acc: 1.0000\n",
      "Epoch 38/50\n",
      "854/854 [==============================] - 60s 71ms/step - loss: 0.0060 - acc: 0.9988 - val_loss: 2.9997e-04 - val_acc: 1.0000\n",
      "Epoch 39/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0086 - acc: 0.9988 - val_loss: 0.0227 - val_acc: 0.9906\n",
      "Epoch 40/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0207 - acc: 0.9941 - val_loss: 4.5355e-04 - val_acc: 1.0000\n",
      "Epoch 41/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0037 - acc: 1.0000 - val_loss: 2.2966e-04 - val_acc: 1.0000\n",
      "Epoch 42/50\n",
      "854/854 [==============================] - 61s 71ms/step - loss: 0.0099 - acc: 0.9965 - val_loss: 0.0209 - val_acc: 0.9906\n",
      "Epoch 43/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0210 - acc: 0.9918 - val_loss: 5.0738e-04 - val_acc: 1.0000\n",
      "Epoch 44/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0059 - acc: 1.0000 - val_loss: 2.5032e-04 - val_acc: 1.0000\n",
      "Epoch 45/50\n",
      "854/854 [==============================] - 66s 77ms/step - loss: 0.0056 - acc: 0.9988 - val_loss: 2.4641e-04 - val_acc: 1.0000\n",
      "Epoch 46/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0062 - acc: 0.9988 - val_loss: 3.6094e-04 - val_acc: 1.0000\n",
      "Epoch 47/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0068 - acc: 1.0000 - val_loss: 1.9959e-04 - val_acc: 1.0000\n",
      "Epoch 48/50\n",
      "854/854 [==============================] - 59s 69ms/step - loss: 0.0035 - acc: 1.0000 - val_loss: 2.2871e-04 - val_acc: 1.0000\n",
      "Epoch 49/50\n",
      "854/854 [==============================] - 60s 70ms/step - loss: 0.0066 - acc: 0.9988 - val_loss: 1.8755e-04 - val_acc: 1.0000\n",
      "Epoch 50/50\n",
      "854/854 [==============================] - 58s 68ms/step - loss: 0.0066 - acc: 0.9965 - val_loss: 2.0213e-04 - val_acc: 1.0000\n",
      "212/212 [==============================] - 6s 28ms/step\n",
      "Loss: 0.15168091220892316\n",
      "Accuracy: 0.9476544919446489\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import keras\n",
    "import sys\n",
    "\n",
    "sys.path.append('/home/inertfluid/.config/spyder-py3/')\n",
    "\n",
    "#from DataPreProcess import LoadData\n",
    "#from ModelCompiler import CompileModel\n",
    "\n",
    "DataSet = 'KinFaceW-I'\n",
    "model = CompileModel()\n",
    "X_Train, Y_Train, X_Test, Y_Test = LoadData()\n",
    "\n",
    "score=[]\n",
    "for j in range(0, 5):\n",
    "    X_Train[j] = np.reshape(list(X_Train[j]), (X_Train[j].shape[0], 64, 64, 6))\n",
    "    Y_Train[j] = keras.utils.to_categorical(Y_Train[j], 2)\n",
    "    X_Test[j] = np.reshape(list(X_Test[j]), (X_Test[j].shape[0], 64, 64, 6))\n",
    "    Y_Test[j] = keras.utils.to_categorical(Y_Test[j], 2)\n",
    "    model.fit(X_Train[j], Y_Train[j], batch_size=64, epochs=50, validation_data=(X_Test[j], Y_Test[j]), shuffle=True)\n",
    "    score+=[model.evaluate(X_Test[j], Y_Test[j], verbose=1)]\n",
    "    \n",
    "score=np.array(score)\n",
    "loss = np.average(score.T[0])\n",
    "accuracy = np.average(score.T[1])\n",
    "\n",
    "print('Loss:', loss)\n",
    "print('Accuracy:', accuracy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[7.19048262e-01, 7.47706423e-01],\n",
       "       [3.67316140e-02, 9.90566037e-01],\n",
       "       [1.90681015e-03, 1.00000000e+00],\n",
       "       [5.15741275e-04, 1.00000000e+00],\n",
       "       [2.02133383e-04, 1.00000000e+00]])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AxesImage(54,36;334.8x217.44)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import cv2\n",
    "from matplotlib import pyplot as plt\n",
    "image1=cv2.imread(\"fd_001_1.jpg\")\n",
    "image2=cv2.imread(\"fd_001_2.jpg\")\n",
    "father = np.concatenate((image1, image2), axis=2)\n",
    "print(plt.imshow(image1))\n",
    "\n",
    "#father.shape\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "AxesImage(54,36;334.8x217.44)\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(plt.imshow(image2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[array([[1]], dtype=uint8) array([[1]], dtype=uint8)\n",
      "  array(['fd_024_1.jpg'], dtype='<U12')\n",
      "  array(['fd_024_2.jpg'], dtype='<U12')]\n",
      " [array([[1]], dtype=uint8) array([[1]], dtype=uint8)\n",
      "  array(['fd_092_1.jpg'], dtype='<U12')\n",
      "  array(['fd_092_2.jpg'], dtype='<U12')]\n",
      " [array([[1]], dtype=uint8) array([[1]], dtype=uint8)\n",
      "  array(['fd_124_1.jpg'], dtype='<U12')\n",
      "  array(['fd_124_2.jpg'], dtype='<U12')]\n",
      " [array([[1]], dtype=uint8) array([[1]], dtype=uint8)\n",
      "  array(['fd_031_1.jpg'], dtype='<U12')\n",
      "  array(['fd_031_2.jpg'], dtype='<U12')]\n",
      " [array([[1]], dtype=uint8) array([[1]], dtype=uint8)\n",
      "  array(['fd_008_1.jpg'], dtype='<U12')\n",
      "  array(['fd_008_2.jpg'], dtype='<U12')]]\n"
     ]
    }
   ],
   "source": [
    "import scipy.io\n",
    "mat_fd = scipy.io.loadmat( 'fd_pairs.mat')\n",
    "print(mat_fd[\"pairs\"][:5,:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.51766187 0.48233816]]\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "father_batch=np.expand_dims(father,axis=0)\n",
    "father_conv=model.predict(father_batch)\n",
    "print(father_conv)\n",
    "#def visualise(father_batch):\n",
    "#    father=np.squeeze(father_batch,axis=0)\n",
    "#    plt.imshow(father)\n",
    "#visualise(father_conv)"
   ]
  }
 ],
 "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": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}