feat:CNN Model

Emotion_Video/emotion_classification_cnn.py

+{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"code","source":"# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:29:42.213352Z\",\"iopub.execute_input\":\"2022-01-02T18:29:42.213653Z\",\"iopub.status.idle\":\"2022-01-02T18:29:47.460037Z\",\"shell.execute_reply.started\":\"2022-01-02T18:29:42.213574Z\",\"shell.execute_reply\":\"2022-01-02T18:29:47.459214Z\"}}\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport pandas as pd\nimport seaborn as sns\nimport os\n\n# Importing Deep Learning Libraries\n\nfrom keras.preprocessing.image import load_img, img_to_array\nfrom keras.preprocessing.image import ImageDataGenerator\nfrom keras.layers import Dense,Input,Dropout,GlobalAveragePooling2D,Flatten,Conv2D,BatchNormalization,Activation,MaxPooling2D\nfrom keras.models import Model,Sequential\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:31:26.874049Z\",\"iopub.execute_input\":\"2022-01-02T18:31:26.87433Z\",\"iopub.status.idle\":\"2022-01-02T18:31:27.174872Z\",\"shell.execute_reply.started\":\"2022-01-02T18:31:26.874299Z\",\"shell.execute_reply\":\"2022-01-02T18:31:27.17414Z\"}}\nfrom tensorflow.keras.optimizers import Adam,SGD,RMSprop\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:31:32.493872Z\",\"iopub.execute_input\":\"2022-01-02T18:31:32.494434Z\",\"iopub.status.idle\":\"2022-01-02T18:31:32.498665Z\",\"shell.execute_reply.started\":\"2022-01-02T18:31:32.494395Z\",\"shell.execute_reply\":\"2022-01-02T18:31:32.497919Z\"}}\npicture_size = 48\nfolder_path = \"../input/face-expression-recognition-dataset/images/\"\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:34:59.744327Z\",\"iopub.execute_input\":\"2022-01-02T18:34:59.744602Z\",\"iopub.status.idle\":\"2022-01-02T18:35:00.78588Z\",\"shell.execute_reply.started\":\"2022-01-02T18:34:59.744566Z\",\"shell.execute_reply\":\"2022-01-02T18:35:00.783248Z\"}}\nexpression = 'disgust'\n\nplt.figure(figsize= (12,12))\nfor i in range(1, 10, 1):\n    plt.subplot(3,3,i)\n    img = load_img(folder_path+\"train/\"+expression+\"/\"+\n                  os.listdir(folder_path + \"train/\" + expression)[i], target_size=(picture_size, picture_size))\n    plt.imshow(img)   \nplt.show()\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:35:05.654001Z\",\"iopub.execute_input\":\"2022-01-02T18:35:05.65435Z\",\"iopub.status.idle\":\"2022-01-02T18:35:21.720425Z\",\"shell.execute_reply.started\":\"2022-01-02T18:35:05.654313Z\",\"shell.execute_reply\":\"2022-01-02T18:35:21.719017Z\"}}\nbatch_size  = 128\n\ndatagen_train  = ImageDataGenerator()\ndatagen_val = ImageDataGenerator()\n\ntrain_set = datagen_train.flow_from_directory(folder_path+\"train\",\n                                              target_size = (picture_size,picture_size),\n                                              color_mode = \"grayscale\",\n                                              batch_size=batch_size,\n                                              class_mode='categorical',\n                                              shuffle=True)\n\n\ntest_set = datagen_val.flow_from_directory(folder_path+\"validation\",\n                                              target_size = (picture_size,picture_size),\n                                              color_mode = \"grayscale\",\n                                              batch_size=batch_size,\n                                              class_mode='categorical',\n                                              shuffle=False)\n\n# %% [markdown]\n# Model Building\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:35:59.134798Z\",\"iopub.execute_input\":\"2022-01-02T18:35:59.135349Z\",\"iopub.status.idle\":\"2022-01-02T18:36:01.780141Z\",\"shell.execute_reply.started\":\"2022-01-02T18:35:59.135309Z\",\"shell.execute_reply\":\"2022-01-02T18:36:01.77949Z\"}}\nno_of_classes = 7\n\nmodel = Sequential()\n\n#1st CNN layer\nmodel.add(Conv2D(64,(3,3),padding = 'same',input_shape = (48,48,1)))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(MaxPooling2D(pool_size = (2,2)))\nmodel.add(Dropout(0.25))\n\n#2nd CNN layer\nmodel.add(Conv2D(128,(5,5),padding = 'same'))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(MaxPooling2D(pool_size = (2,2)))\nmodel.add(Dropout (0.25))\n\n#3rd CNN layer\nmodel.add(Conv2D(512,(3,3),padding = 'same'))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(MaxPooling2D(pool_size = (2,2)))\nmodel.add(Dropout (0.25))\n\n#4th CNN layer\nmodel.add(Conv2D(512,(3,3), padding='same'))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(MaxPooling2D(pool_size=(2, 2)))\nmodel.add(Dropout(0.25))\n\nmodel.add(Flatten())\n\n#Fully connected 1st layer\nmodel.add(Dense(256))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.25))\n\n\n# Fully connected layer 2nd layer\nmodel.add(Dense(512))\nmodel.add(BatchNormalization())\nmodel.add(Activation('relu'))\nmodel.add(Dropout(0.25))\n\nmodel.add(Dense(no_of_classes, activation='softmax'))\n\n\n\nopt = Adam(lr = 0.0001)\nmodel.compile(optimizer=opt,loss='categorical_crossentropy', metrics=['accuracy'])\nmodel.summary()\n\n# %% [markdown]\n# Fitting the Model with Training and Validation Data\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:36:45.963735Z\",\"iopub.execute_input\":\"2022-01-02T18:36:45.963996Z\",\"iopub.status.idle\":\"2022-01-02T18:36:45.978165Z\",\"shell.execute_reply.started\":\"2022-01-02T18:36:45.963967Z\",\"shell.execute_reply\":\"2022-01-02T18:36:45.977455Z\"}}\nfrom keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau\n\ncheckpoint = ModelCheckpoint(\"./model.h5\", monitor='val_acc', verbose=1, save_best_only=True, mode='max')\n\nearly_stopping = EarlyStopping(monitor='val_loss',\n                          min_delta=0,\n                          patience=3,\n                          verbose=1,\n                          restore_best_weights=True\n                          )\n\nreduce_learningrate = ReduceLROnPlateau(monitor='val_loss',\n                              factor=0.2,\n                              patience=3,\n                              verbose=1,\n                              min_delta=0.0001)\n\ncallbacks_list = [early_stopping,checkpoint,reduce_learningrate]\n\nepochs = 48\n\nmodel.compile(loss='categorical_crossentropy',\n              optimizer = Adam(lr=0.001),\n              metrics=['accuracy'])\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:37:22.724599Z\",\"iopub.execute_input\":\"2022-01-02T18:37:22.724884Z\",\"iopub.status.idle\":\"2022-01-02T18:47:27.839479Z\",\"shell.execute_reply.started\":\"2022-01-02T18:37:22.724853Z\",\"shell.execute_reply\":\"2022-01-02T18:47:27.838695Z\"}}\nhistory = model.fit_generator(generator=train_set,\n                                steps_per_epoch=train_set.n//train_set.batch_size,\n                                epochs=epochs,\n                                validation_data = test_set,\n                                validation_steps = test_set.n//test_set.batch_size,\n                                callbacks=callbacks_list\n                                )\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T18:52:03.305144Z\",\"iopub.execute_input\":\"2022-01-02T18:52:03.305869Z\",\"iopub.status.idle\":\"2022-01-02T18:52:27.411667Z\",\"shell.execute_reply.started\":\"2022-01-02T18:52:03.305831Z\",\"shell.execute_reply\":\"2022-01-02T18:52:27.410996Z\"}}\nplt.style.use('dark_background')\n\nplt.figure(figsize=(20,10))\nplt.subplot(1, 2, 1)\nplt.suptitle('Optimizer : Adam', fontsize=10)\nplt.ylabel('Loss', fontsize=16)\nplt.plot(history.history['loss'], label='Training Loss')\nplt.plot(history.history['val_loss'], label='Validation Loss')\nplt.legend(loc='upper right')\n\nplt.subplot(1, 2, 2)\nplt.ylabel('Accuracy', fontsize=16)\nplt.plot(history.history['accuracy'], label='Training Accuracy')\nplt.plot(history.history['val_accuracy'], label='Validation Accuracy')\nplt.legend(loc='lower right')\nplt.show()\n\n# %% [code] {\"execution\":{\"iopub.status.busy\":\"2022-01-02T19:08:54.788074Z\",\"iopub.execute_input\":\"2022-01-02T19:08:54.788346Z\",\"iopub.status.idle\":\"2022-01-02T19:08:54.950179Z\",\"shell.execute_reply.started\":\"2022-01-02T19:08:54.788317Z\",\"shell.execute_reply\":\"2022-01-02T19:08:54.949443Z\"}}\nmodel.save('/kaggle/working/final_model.h5')","metadata":{"_uuid":"f7dbf1bc-5f64-4f46-a2c9-0831213bb3b1","_cell_guid":"80c6e78c-ee66-49d7-bae1-88859794c451","collapsed":false,"jupyter":{"outputs_hidden":false},"trusted":true},"execution_count":null,"outputs":[]}]}
\ No newline at end of file