cnn_arch_2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import data_utils as my_utils\n",
    "import cv2\n",
    "import dlib\n",
    "import itertools\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "import keras\n",
    "from keras.callbacks import ModelCheckpoint\n",
    "from keras.layers import Activation, BatchNormalization, Conv2D, Dense, Dropout, Flatten, Input, MaxPool2D\n",
    "from keras.models import Sequential, Model\n",
    "from keras.utils import np_utils\n",
    "from skimage.feature import hog\n",
    "import sklearn\n",
    "from sklearn.metrics import confusion_matrix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"fer2013.csv\")\n",
    "emotion_labels = {0: \"Angry\", 1: \"Disgust\", 2:\"Fear\", 3:\"Happy\", 4:\"Sad\", 5:\"Surprise\", 6:\"Neutral\"}\n",
    "labels = [\"Angry\", \"Disgust\", \"Fear\", \"Happy\", \"Sad\", \"Surprise\", \"Neutral\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>emotion</th>\n",
       "      <th>pixels</th>\n",
       "      <th>Usage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>70 80 82 72 58 58 60 63 54 58 60 48 89 115 121...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>151 150 147 155 148 133 111 140 170 174 182 15...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>231 212 156 164 174 138 161 173 182 200 106 38...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>24 32 36 30 32 23 19 20 30 41 21 22 32 34 21 1...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6</td>\n",
       "      <td>4 0 0 0 0 0 0 0 0 0 0 0 3 15 23 28 48 50 58 84...</td>\n",
       "      <td>Training</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   emotion                                             pixels     Usage\n",
       "0        0  70 80 82 72 58 58 60 63 54 58 60 48 89 115 121...  Training\n",
       "1        0  151 150 147 155 148 133 111 140 170 174 182 15...  Training\n",
       "2        2  231 212 156 164 174 138 161 173 182 200 106 38...  Training\n",
       "3        4  24 32 36 30 32 23 19 20 30 41 21 22 32 34 21 1...  Training\n",
       "4        6  4 0 0 0 0 0 0 0 0 0 0 0 3 15 23 28 48 50 58 84...  Training"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Training       28709\n",
       "PrivateTest     3589\n",
       "PublicTest      3589\n",
       "Name: Usage, dtype: int64"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"Usage\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "3    8989\n",
       "6    6198\n",
       "4    6077\n",
       "2    5121\n",
       "0    4953\n",
       "5    4002\n",
       "1     547\n",
       "Name: emotion, dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[\"emotion\"].value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{0: 'Angry',\n",
       " 1: 'Disgust',\n",
       " 2: 'Fear',\n",
       " 3: 'Happy',\n",
       " 4: 'Sad',\n",
       " 5: 'Surprise',\n",
       " 6: 'Neutral'}"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "emotion_labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/Users/ritvikareddy 1/Documents/GitHub/ml-project/venv/lib/python3.6/site-packages/pandas/core/frame.py:3697: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  errors=errors)\n",
      "/Users/ritvikareddy 1/Documents/GitHub/ml-project/data_utils.py:19: SettingWithCopyWarning: \n",
      "A value is trying to be set on a copy of a slice from a DataFrame.\n",
      "Try using .loc[row_indexer,col_indexer] = value instead\n",
      "\n",
      "See the caveats in the documentation: http://pandas.pydata.org/pandas-docs/stable/indexing.html#indexing-view-versus-copy\n",
      "  lambda pixel_str: np.fromstring(pixel_str, sep=' '))\n"
     ]
    }
   ],
   "source": [
    "# process the training, validation and test data\n",
    "x_train, y_train = my_utils.process_data(df, 'Training')\n",
    "x_validation, y_validation = my_utils.process_data(df, 'PublicTest')\n",
    "x_test, y_test = my_utils.process_data(df, \"PrivateTest\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# get face landmarks\n",
    "x_train_landmarks = my_utils.extract_landmark_features(x_train)\n",
    "x_validation_landmarks = my_utils.extract_landmark_features(x_validation)\n",
    "x_test_landmarks = my_utils.extract_landmark_features(x_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_1 (InputLayer)            (None, 48, 48, 1)    0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_1 (Conv2D)               (None, 46, 46, 64)   640         input_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_1 (BatchNor (None, 46, 46, 64)   256         conv2d_1[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_1 (MaxPooling2D)  (None, 22, 22, 64)   0           batch_normalization_1[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_2 (Conv2D)               (None, 20, 20, 128)  73856       max_pooling2d_1[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_2 (BatchNor (None, 20, 20, 128)  512         conv2d_2[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_2 (MaxPooling2D)  (None, 9, 9, 128)    0           batch_normalization_2[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_3 (Conv2D)               (None, 7, 7, 256)    295168      max_pooling2d_2[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_3 (BatchNor (None, 7, 7, 256)    1024        conv2d_3[0][0]                   \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_3 (MaxPooling2D)  (None, 3, 3, 256)    0           batch_normalization_3[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "dropout_1 (Dropout)             (None, 3, 3, 256)    0           max_pooling2d_3[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "flatten_1 (Flatten)             (None, 2304)         0           dropout_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "input_2 (InputLayer)            (None, 68, 2)        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "dense_1 (Dense)                 (None, 4096)         9441280     flatten_1[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "flatten_2 (Flatten)             (None, 136)          0           input_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dropout_2 (Dropout)             (None, 4096)         0           dense_1[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_4 (Dense)                 (None, 1024)         140288      flatten_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 1024)         4195328     dropout_2[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_5 (BatchNor (None, 1024)         4096        dense_4[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_4 (BatchNor (None, 1024)         4096        dense_2[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "dense_5 (Dense)                 (None, 128)          131200      batch_normalization_5[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "dense_3 (Dense)                 (None, 128)          131200      batch_normalization_4[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_6 (BatchNor (None, 128)          512         dense_5[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_1 (Concatenate)     (None, 256)          0           dense_3[0][0]                    \n",
      "                                                                 batch_normalization_6[0][0]      \n",
      "__________________________________________________________________________________________________\n",
      "dense_6 (Dense)                 (None, 7)            1799        concatenate_1[0][0]              \n",
      "==================================================================================================\n",
      "Total params: 14,421,255\n",
      "Trainable params: 14,416,007\n",
      "Non-trainable params: 5,248\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "# cnn for just the raw pixels\n",
    "# Input layer 1\n",
    "images_nn = Input(shape=(48,48,1))\n",
    "\n",
    "# Layer 1\n",
    "images_conv1 = Conv2D(64, 3, kernel_initializer='he_normal', activation='relu')(images_nn)\n",
    "images_batchnorm1 = BatchNormalization()(images_conv1)\n",
    "images_pool1 = MaxPool2D(pool_size=(3,3), strides=2)(images_batchnorm1)\n",
    "\n",
    "# Layer 2\n",
    "images_conv2 = Conv2D(128, 3, activation='relu')(images_pool1)\n",
    "images_batchnorm2 = BatchNormalization()(images_conv2)\n",
    "images_pool2 = MaxPool2D(pool_size=(3,3), strides=2)(images_batchnorm2)\n",
    "\n",
    "# Layer 3\n",
    "images_conv3 = Conv2D(256, 3, activation='relu')(images_pool2)\n",
    "images_batchnorm3 = BatchNormalization()(images_conv3)\n",
    "images_pool3 = MaxPool2D(pool_size=(3,3), strides=2)(images_batchnorm3)\n",
    "images_dropout1 = Dropout(0.6)(images_pool3)\n",
    "\n",
    "# Flatten\n",
    "images_flat1 = Flatten()(images_dropout1)\n",
    "\n",
    "# Fully Connected Layer 1\n",
    "images_dense1 = Dense(4096, activation='relu')(images_flat1)\n",
    "images_dropout2 = Dropout(0.6)(images_dense1)\n",
    "\n",
    "# Fully Connected Layer 2\n",
    "images_dense2 = Dense(1024, activation='relu')(images_dropout2)\n",
    "images_batchnorm4 = BatchNormalization()(images_dense2)\n",
    "\n",
    "# Fully Connected Layer 3\n",
    "images_dense3 = Dense(128, activation='relu')(images_batchnorm4)\n",
    "\n",
    "\n",
    "# cnn for landmark features\n",
    "# input shape will be 68*2 (from landamrks)\n",
    "# Input Layer\n",
    "landmarks_features_nn = Input(shape=(68,2))\n",
    "\n",
    "# Flatten\n",
    "landmarks_features_flat1 = Flatten()(landmarks_features_nn)\n",
    "\n",
    "# Fully Connected Layer 4\n",
    "landmarks_features_dense1 = Dense(1024, activation='relu')(landmarks_features_flat1)\n",
    "landmarks_features_batchnorm1 = BatchNormalization()(landmarks_features_dense1)\n",
    "\n",
    "# Fully Connected Layer 5\n",
    "landmarks_features_dense2 = Dense(128, activation='relu') (landmarks_features_batchnorm1)\n",
    "landmarks_features_batchnorm2 = BatchNormalization()(landmarks_features_dense2)\n",
    "\n",
    "# merge final outputs from both networks\n",
    "merged_net = keras.layers.merge.concatenate([images_dense3, landmarks_features_batchnorm2],\n",
    "                                           axis=-1)\n",
    "\n",
    "# Output Layer\n",
    "output_layer = Dense(7, activation='softmax')(merged_net)\n",
    "\n",
    "model = Model(inputs=[images_nn, landmarks_features_nn], outputs=output_layer)\n",
    "model.compile(loss='categorical_crossentropy', optimizer='adam', metrics=['accuracy'])\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train on 28709 samples, validate on 3589 samples\n",
      "Epoch 1/10\n",
      " - 465s - loss: 1.5616 - acc: 0.4233 - val_loss: 1.9612 - val_acc: 0.3586\n",
      "\n",
      "Epoch 00001: val_loss improved from inf to 1.96117, saving model to cnn_arch2_landmarks\n",
      "Epoch 2/10\n",
      " - 472s - loss: 1.3852 - acc: 0.4719 - val_loss: 1.8763 - val_acc: 0.3695\n",
      "\n",
      "Epoch 00002: val_loss improved from 1.96117 to 1.87634, saving model to cnn_arch2_landmarks\n",
      "Epoch 3/10\n",
      " - 529s - loss: 1.3196 - acc: 0.4980 - val_loss: 3.2307 - val_acc: 0.2026\n",
      "\n",
      "Epoch 00003: val_loss did not improve from 1.87634\n",
      "Epoch 4/10\n",
      " - 541s - loss: 1.2571 - acc: 0.5226 - val_loss: 1.5528 - val_acc: 0.4154\n",
      "\n",
      "Epoch 00004: val_loss improved from 1.87634 to 1.55283, saving model to cnn_arch2_landmarks\n",
      "Epoch 5/10\n",
      " - 522s - loss: 1.2371 - acc: 0.5288 - val_loss: 1.8315 - val_acc: 0.3511\n",
      "\n",
      "Epoch 00005: val_loss did not improve from 1.55283\n",
      "Epoch 6/10\n",
      " - 514s - loss: 1.1722 - acc: 0.5538 - val_loss: 1.8235 - val_acc: 0.3915\n",
      "\n",
      "Epoch 00006: val_loss did not improve from 1.55283\n",
      "Epoch 7/10\n",
      " - 516s - loss: 1.1282 - acc: 0.5739 - val_loss: 1.4518 - val_acc: 0.4812\n",
      "\n",
      "Epoch 00007: val_loss improved from 1.55283 to 1.45179, saving model to cnn_arch2_landmarks\n",
      "Epoch 8/10\n",
      " - 581s - loss: 1.0993 - acc: 0.5847 - val_loss: 1.3136 - val_acc: 0.4990\n",
      "\n",
      "Epoch 00008: val_loss improved from 1.45179 to 1.31363, saving model to cnn_arch2_landmarks\n",
      "Epoch 9/10\n",
      " - 554s - loss: 1.0612 - acc: 0.5985 - val_loss: 1.7037 - val_acc: 0.4196\n",
      "\n",
      "Epoch 00009: val_loss did not improve from 1.31363\n",
      "Epoch 10/10\n",
      " - 521s - loss: 1.0247 - acc: 0.6145 - val_loss: 1.2070 - val_acc: 0.5453\n",
      "\n",
      "Epoch 00010: val_loss improved from 1.31363 to 1.20699, saving model to cnn_arch2_landmarks\n"
     ]
    }
   ],
   "source": [
    "# save best weights\n",
    "checkpoint = ModelCheckpoint(filepath='cnn_arch2_landmarks', verbose=2, save_best_only=True)\n",
    "\n",
    "# num epochs\n",
    "epochs = 10\n",
    "\n",
    "# run model\n",
    "model_history = model.fit([x_train, x_train_landmarks], y_train, epochs=epochs,\n",
    "                 shuffle=True,\n",
    "                 batch_size=100, validation_data=([x_validation, x_validation_landmarks], y_validation),\n",
    "                 callbacks=[checkpoint], verbose=2)\n",
    "\n",
    "# save model to json\n",
    "model_json = model.to_json()\n",
    "with open(\"cnn_arch2_landmarks.json\", \"w\") as json_file:\n",
    "    json_file.write(model_json)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.figure(figsize=(12,4))\n",
    "plt.subplot(1, 2, 1)\n",
    "plt.suptitle('CNN on raw pixel data', fontsize=10)\n",
    "plt.ylabel('Loss', fontsize=16)\n",
    "plt.plot(model_history.history['loss'], color='b', label='Training Loss')\n",
    "plt.plot(model_history.history['val_loss'], color='r', label='Validation Loss')\n",
    "plt.legend(loc='upper right')\n",
    "\n",
    "plt.subplot(1, 2, 2)\n",
    "plt.ylabel('Accuracy', fontsize=16)\n",
    "plt.plot(model_history.history['acc'], color='b', label='Training Accuracy')\n",
    "plt.plot(model_history.history['val_acc'], color='r', label='Validation Accuracy')\n",
    "plt.legend(loc='lower right')\n",
    "plt.show()\n",
    "fig.savefig('Loss_comparison_model_arch2.png')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3589/3589 [==============================] - 15s 4ms/step\n"
     ]
    }
   ],
   "source": [
    "y_predicted_probs = model.predict([x_test, x_test_landmarks], verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3589/3589 [==============================] - 14s 4ms/step\n"
     ]
    }
   ],
   "source": [
    "score = model.evaluate([x_test, x_test_landmarks], y_test, verbose=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Metrics:  ['loss', 'acc']\n",
      "Loss on test data 1.1701981891460425\n",
      "Test Accuracy 0.5583728058287014\n"
     ]
    }
   ],
   "source": [
    "print(\"Metrics: \", model.metrics_names)\n",
    "print(\"Loss on test data\", score[0])\n",
    "print(\"Test Accuracy\", score[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_predicted_labels = my_utils.convert_probs_to_labels(y_predicted_probs, labels)\n",
    "y_actual_labels = my_utils.convert_probs_to_labels(y_test, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Confusion matrix, without normalization\n",
      "[[365   2   7  33  13  54  17]\n",
      " [ 41   6   0   3   0   3   2]\n",
      " [170   0  81  32  17 137  91]\n",
      " [ 48   0  13 730  19  51  18]\n",
      " [189   0   8  58 206 157   8]\n",
      " [170   0  23  49  39 305   8]\n",
      " [ 32   0  17  35   8  13 311]]\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "cnf_matrix = sklearn.metrics.confusion_matrix(y_actual_labels, y_predicted_labels)\n",
    "np.set_printoptions(precision=2)\n",
    "\n",
    "# Plot non-normalized confusion matrix\n",
    "fig = my_utils.plot_confusion_matrix(cnf_matrix, classes=labels,\n",
    "                                     title='Confusion Matrix for Test Data')\n",
    "plt.show()\n",
    "fig.savefig('confusion_matrix_arch2.png')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.save('model_arch2.h5')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "keras.utils.plot_model(model, to_file='model_arch2.png', show_shapes=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test_for_single_image(file_name, model, labels):\n",
    "    image_pixels = cv2.imread(file_name, 0)\n",
    "#     print(len(image_pixels))\n",
    "    image_input = image_pixels.reshape(-1, 48, 48, 1)\n",
    "    predicted_emotion = model.predict(image_input, verbose=0)\n",
    "#     print(predicted_emotion)\n",
    "    print(\"This image is classified to be {} with {:.2f}% confidence\".format(\n",
    "        labels[predicted_emotion.argmax()],predicted_emotion.max()*100))\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('100th.jpg', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('ashok.jpg', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('ashok2.png', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('rose2.jpg', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('rose3.jpg', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('jose.png', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('rohit_happy.jpeg', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_for_single_image('ankita.png', model, labels)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}