VGG16_Modified/vgg16-modified-fer.ipynb

{"cells":[{"metadata":{"_uuid":"d629ff2d2480ee46fbb7e2d37f6b5fab8052498a","_cell_guid":"79c7e3d0-c299-4dcb-8224-4455121ee9b0","trusted":true},"cell_type":"code","source":"import numpy as np\nimport tensorflow as tf\nimport matplotlib.pyplot as plt\nimport pandas as pd\nimport os\nfrom tensorflow.keras.preprocessing.image import ImageDataGenerator","execution_count":1,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"train_datagen = ImageDataGenerator(rescale = 1./255,rotation_range=15,\n                                    width_shift_range=0.15,\n                                    height_shift_range=0.15,\n                                    shear_range=0.15,\n                                    zoom_range=0.15,\n                                    horizontal_flip=True,\n                                  )\nval_datagen = ImageDataGenerator(rescale=1./255)","execution_count":2,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"#Batch Size and Image Size\nbatch_sz = 64\nsz = 48\n\ntrain_generator = train_datagen.flow_from_directory('../input/fer-dsc-v3/final_data_v5/final_data_v5/Train',\n                                                    batch_size = batch_sz,\n                                                    class_mode='categorical',\n                                                    target_size=(sz,sz))\n\nval_generator = val_datagen.flow_from_directory('../input/fer-dsc-v3/final_data_v5/final_data_v5/Test',\n                                                    batch_size = batch_sz,\n                                                    class_mode='categorical',\n                                                    target_size=(sz,sz))","execution_count":3,"outputs":[{"output_type":"stream","text":"Found 13651 images belonging to 4 classes.\nFound 1649 images belonging to 4 classes.\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.applications.vgg16 import VGG16","execution_count":4,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"pre_model = VGG16(include_top = False,\n                 weights = 'imagenet',\n                 input_shape = (48,48,3))","execution_count":5,"outputs":[{"output_type":"stream","text":"Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/vgg16/vgg16_weights_tf_dim_ordering_tf_kernels_notop.h5\n58892288/58889256 [==============================] - 0s 0us/step\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"for layer in pre_model.layers:\n    layer.trainable=False","execution_count":6,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"last_output = pre_model.get_layer('block5_conv1').output","execution_count":7,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"x = tf.keras.layers.GlobalAveragePooling2D()(last_output)\nx = tf.keras.layers.Flatten()(last_output)\nx = tf.keras.layers.Dense(256,activation='relu')(x)\nx = tf.keras.layers.Dropout(0.4)(x)\nx = tf.keras.layers.Dense(128,activation= 'relu')(x)\nx = tf.keras.layers.Dropout(0.2)(x)\nx = tf.keras.layers.Dense(4,activation='softmax')(x)","execution_count":8,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model = tf.keras.Model(pre_model.input, x)","execution_count":9,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"opt = tf.keras.optimizers.RMSprop(learning_rate=1e-5)\nmodel.compile(optimizer=opt,loss='categorical_crossentropy',metrics=['accuracy'])","execution_count":10,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"history = model.fit(train_generator,\n                   validation_data=val_generator,\n                    steps_per_epoch=train_generator.samples//train_generator.batch_size,\n                   validation_steps=val_generator.samples//val_generator.batch_size,\n                    epochs=30\n                    )","execution_count":11,"outputs":[{"output_type":"stream","text":"Epoch 1/30\n213/213 [==============================] - 42s 195ms/step - loss: 1.7592 - accuracy: 0.2820 - val_loss: 1.3649 - val_accuracy: 0.3244\nEpoch 2/30\n213/213 [==============================] - 23s 107ms/step - loss: 1.4988 - accuracy: 0.2955 - val_loss: 1.3303 - val_accuracy: 0.3444\nEpoch 3/30\n213/213 [==============================] - 23s 107ms/step - loss: 1.4096 - accuracy: 0.3199 - val_loss: 1.3097 - val_accuracy: 0.3862\nEpoch 4/30\n213/213 [==============================] - 24s 111ms/step - loss: 1.3798 - accuracy: 0.3358 - val_loss: 1.2957 - val_accuracy: 0.4056\nEpoch 5/30\n213/213 [==============================] - 22s 104ms/step - loss: 1.3629 - accuracy: 0.3428 - val_loss: 1.2850 - val_accuracy: 0.4275\nEpoch 6/30\n213/213 [==============================] - 23s 106ms/step - loss: 1.3568 - accuracy: 0.3469 - val_loss: 1.2658 - val_accuracy: 0.4275\nEpoch 7/30\n213/213 [==============================] - 22s 103ms/step - loss: 1.3352 - accuracy: 0.3660 - val_loss: 1.2582 - val_accuracy: 0.4387\nEpoch 8/30\n213/213 [==============================] - 22s 104ms/step - loss: 1.3228 - accuracy: 0.3768 - val_loss: 1.2499 - val_accuracy: 0.4550\nEpoch 9/30\n213/213 [==============================] - 23s 107ms/step - loss: 1.3237 - accuracy: 0.3795 - val_loss: 1.2392 - val_accuracy: 0.4544\nEpoch 10/30\n213/213 [==============================] - 23s 107ms/step - loss: 1.3168 - accuracy: 0.3823 - val_loss: 1.2352 - val_accuracy: 0.4544\nEpoch 11/30\n213/213 [==============================] - 24s 111ms/step - loss: 1.3107 - accuracy: 0.3876 - val_loss: 1.2281 - val_accuracy: 0.4700\nEpoch 12/30\n213/213 [==============================] - 22s 105ms/step - loss: 1.2975 - accuracy: 0.3936 - val_loss: 1.2213 - val_accuracy: 0.4712\nEpoch 13/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2957 - accuracy: 0.3969 - val_loss: 1.2142 - val_accuracy: 0.4675\nEpoch 14/30\n213/213 [==============================] - 23s 106ms/step - loss: 1.2875 - accuracy: 0.4013 - val_loss: 1.2018 - val_accuracy: 0.4775\nEpoch 15/30\n213/213 [==============================] - 22s 103ms/step - loss: 1.2826 - accuracy: 0.4060 - val_loss: 1.1974 - val_accuracy: 0.4787\nEpoch 16/30\n213/213 [==============================] - 22s 105ms/step - loss: 1.2774 - accuracy: 0.4118 - val_loss: 1.1941 - val_accuracy: 0.4775\nEpoch 17/30\n213/213 [==============================] - 23s 106ms/step - loss: 1.2663 - accuracy: 0.4270 - val_loss: 1.1842 - val_accuracy: 0.4812\nEpoch 18/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2708 - accuracy: 0.4158 - val_loss: 1.1803 - val_accuracy: 0.4831\nEpoch 19/30\n213/213 [==============================] - 22s 101ms/step - loss: 1.2645 - accuracy: 0.4238 - val_loss: 1.1766 - val_accuracy: 0.4950\nEpoch 20/30\n213/213 [==============================] - 23s 106ms/step - loss: 1.2667 - accuracy: 0.4256 - val_loss: 1.1747 - val_accuracy: 0.4900\nEpoch 21/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2505 - accuracy: 0.4307 - val_loss: 1.1704 - val_accuracy: 0.4875\nEpoch 22/30\n213/213 [==============================] - 22s 105ms/step - loss: 1.2548 - accuracy: 0.4302 - val_loss: 1.1688 - val_accuracy: 0.4906\nEpoch 23/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2518 - accuracy: 0.4308 - val_loss: 1.1611 - val_accuracy: 0.4969\nEpoch 24/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2414 - accuracy: 0.4384 - val_loss: 1.1502 - val_accuracy: 0.5019\nEpoch 25/30\n213/213 [==============================] - 23s 106ms/step - loss: 1.2408 - accuracy: 0.4392 - val_loss: 1.1507 - val_accuracy: 0.4981\nEpoch 26/30\n213/213 [==============================] - 22s 103ms/step - loss: 1.2399 - accuracy: 0.4454 - val_loss: 1.1418 - val_accuracy: 0.5063\nEpoch 27/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2367 - accuracy: 0.4454 - val_loss: 1.1435 - val_accuracy: 0.4975\nEpoch 28/30\n213/213 [==============================] - 22s 105ms/step - loss: 1.2362 - accuracy: 0.4451 - val_loss: 1.1367 - val_accuracy: 0.5038\nEpoch 29/30\n213/213 [==============================] - 22s 102ms/step - loss: 1.2277 - accuracy: 0.4526 - val_loss: 1.1374 - val_accuracy: 0.5063\nEpoch 30/30\n213/213 [==============================] - 22s 104ms/step - loss: 1.2307 - accuracy: 0.4462 - val_loss: 1.1341 - val_accuracy: 0.5144\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"for layer in model.layers:\n    layer.trainable = True","execution_count":12,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"print(model.summary())","execution_count":13,"outputs":[{"output_type":"stream","text":"Model: \"functional_1\"\n_________________________________________________________________\nLayer (type)                 Output Shape              Param #   \n=================================================================\ninput_1 (InputLayer)         [(None, 48, 48, 3)]       0         \n_________________________________________________________________\nblock1_conv1 (Conv2D)        (None, 48, 48, 64)        1792      \n_________________________________________________________________\nblock1_conv2 (Conv2D)        (None, 48, 48, 64)        36928     \n_________________________________________________________________\nblock1_pool (MaxPooling2D)   (None, 24, 24, 64)        0         \n_________________________________________________________________\nblock2_conv1 (Conv2D)        (None, 24, 24, 128)       73856     \n_________________________________________________________________\nblock2_conv2 (Conv2D)        (None, 24, 24, 128)       147584    \n_________________________________________________________________\nblock2_pool (MaxPooling2D)   (None, 12, 12, 128)       0         \n_________________________________________________________________\nblock3_conv1 (Conv2D)        (None, 12, 12, 256)       295168    \n_________________________________________________________________\nblock3_conv2 (Conv2D)        (None, 12, 12, 256)       590080    \n_________________________________________________________________\nblock3_conv3 (Conv2D)        (None, 12, 12, 256)       590080    \n_________________________________________________________________\nblock3_pool (MaxPooling2D)   (None, 6, 6, 256)         0         \n_________________________________________________________________\nblock4_conv1 (Conv2D)        (None, 6, 6, 512)         1180160   \n_________________________________________________________________\nblock4_conv2 (Conv2D)        (None, 6, 6, 512)         2359808   \n_________________________________________________________________\nblock4_conv3 (Conv2D)        (None, 6, 6, 512)         2359808   \n_________________________________________________________________\nblock4_pool (MaxPooling2D)   (None, 3, 3, 512)         0         \n_________________________________________________________________\nblock5_conv1 (Conv2D)        (None, 3, 3, 512)         2359808   \n_________________________________________________________________\nflatten (Flatten)            (None, 4608)              0         \n_________________________________________________________________\ndense (Dense)                (None, 256)               1179904   \n_________________________________________________________________\ndropout (Dropout)            (None, 256)               0         \n_________________________________________________________________\ndense_1 (Dense)              (None, 128)               32896     \n_________________________________________________________________\ndropout_1 (Dropout)          (None, 128)               0         \n_________________________________________________________________\ndense_2 (Dense)              (None, 4)                 516       \n=================================================================\nTotal params: 11,208,388\nTrainable params: 11,208,388\nNon-trainable params: 0\n_________________________________________________________________\nNone\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"opt1 = tf.keras.optimizers.Adam(learning_rate = 1e-4)\nmodel.compile(optimizer = opt1, loss='categorical_crossentropy', metrics=['accuracy'])","execution_count":14,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"history1 = model.fit(train_generator,\n                    validation_data = val_generator,\n                    validation_steps = val_generator.samples // val_generator.batch_size,\n                    epochs = 25)\n                    #callbacks = [callbacks])","execution_count":15,"outputs":[{"output_type":"stream","text":"Epoch 1/25\n214/214 [==============================] - 23s 108ms/step - loss: 1.1612 - accuracy: 0.4804 - val_loss: 0.9332 - val_accuracy: 0.5919\nEpoch 2/25\n214/214 [==============================] - 23s 108ms/step - loss: 0.9568 - accuracy: 0.5912 - val_loss: 0.8796 - val_accuracy: 0.6181\nEpoch 3/25\n214/214 [==============================] - 24s 113ms/step - loss: 0.8838 - accuracy: 0.6327 - val_loss: 0.8063 - val_accuracy: 0.6669\nEpoch 4/25\n214/214 [==============================] - 23s 109ms/step - loss: 0.8293 - accuracy: 0.6614 - val_loss: 0.7916 - val_accuracy: 0.6700\nEpoch 5/25\n214/214 [==============================] - 23s 108ms/step - loss: 0.7907 - accuracy: 0.6819 - val_loss: 0.7363 - val_accuracy: 0.6994\nEpoch 6/25\n214/214 [==============================] - 24s 110ms/step - loss: 0.7672 - accuracy: 0.6877 - val_loss: 0.7258 - val_accuracy: 0.7050\nEpoch 7/25\n214/214 [==============================] - 23s 107ms/step - loss: 0.7431 - accuracy: 0.7044 - val_loss: 0.6983 - val_accuracy: 0.7081\nEpoch 8/25\n214/214 [==============================] - 24s 113ms/step - loss: 0.7057 - accuracy: 0.7185 - val_loss: 0.7255 - val_accuracy: 0.7019\nEpoch 9/25\n214/214 [==============================] - 23s 107ms/step - loss: 0.6936 - accuracy: 0.7252 - val_loss: 0.7029 - val_accuracy: 0.7144\nEpoch 10/25\n214/214 [==============================] - 23s 110ms/step - loss: 0.6680 - accuracy: 0.7357 - val_loss: 0.7118 - val_accuracy: 0.7031\nEpoch 11/25\n214/214 [==============================] - 24s 112ms/step - loss: 0.6477 - accuracy: 0.7420 - val_loss: 0.6681 - val_accuracy: 0.7275\nEpoch 12/25\n214/214 [==============================] - 23s 108ms/step - loss: 0.6266 - accuracy: 0.7520 - val_loss: 0.6707 - val_accuracy: 0.7262\nEpoch 13/25\n214/214 [==============================] - 24s 113ms/step - loss: 0.6078 - accuracy: 0.7599 - val_loss: 0.7592 - val_accuracy: 0.7038\nEpoch 14/25\n214/214 [==============================] - 23s 109ms/step - loss: 0.5854 - accuracy: 0.7676 - val_loss: 0.6814 - val_accuracy: 0.7294\nEpoch 15/25\n214/214 [==============================] - 24s 110ms/step - loss: 0.5645 - accuracy: 0.7776 - val_loss: 0.8102 - val_accuracy: 0.7106\nEpoch 16/25\n214/214 [==============================] - 24s 114ms/step - loss: 0.5530 - accuracy: 0.7840 - val_loss: 0.6473 - val_accuracy: 0.7437\nEpoch 17/25\n214/214 [==============================] - 23s 109ms/step - loss: 0.5274 - accuracy: 0.7950 - val_loss: 0.7323 - val_accuracy: 0.7281\nEpoch 18/25\n214/214 [==============================] - 25s 115ms/step - loss: 0.5070 - accuracy: 0.7996 - val_loss: 0.6749 - val_accuracy: 0.7350\nEpoch 19/25\n214/214 [==============================] - 23s 109ms/step - loss: 0.4891 - accuracy: 0.8101 - val_loss: 0.6697 - val_accuracy: 0.7419\nEpoch 20/25\n214/214 [==============================] - 23s 109ms/step - loss: 0.4740 - accuracy: 0.8170 - val_loss: 0.6967 - val_accuracy: 0.7450\nEpoch 21/25\n214/214 [==============================] - 24s 111ms/step - loss: 0.4582 - accuracy: 0.8193 - val_loss: 0.6966 - val_accuracy: 0.7462\nEpoch 22/25\n214/214 [==============================] - 23s 108ms/step - loss: 0.4422 - accuracy: 0.8281 - val_loss: 0.7082 - val_accuracy: 0.7344\nEpoch 23/25\n214/214 [==============================] - 24s 114ms/step - loss: 0.4238 - accuracy: 0.8371 - val_loss: 0.7213 - val_accuracy: 0.7450\nEpoch 24/25\n214/214 [==============================] - 24s 113ms/step - loss: 0.4078 - accuracy: 0.8384 - val_loss: 0.7147 - val_accuracy: 0.7487\nEpoch 25/25\n214/214 [==============================] - 23s 109ms/step - loss: 0.3863 - accuracy: 0.8514 - val_loss: 0.7504 - val_accuracy: 0.7525\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"N = 25\nplt.style.use(\"ggplot\")\nplt.figure()\nplt.plot(np.arange(0, N), history1.history[\"loss\"], label=\"train_loss\")\nplt.plot(np.arange(0, N), history1.history[\"val_loss\"], label=\"val_loss\")\nplt.plot(np.arange(0, N), history1.history[\"accuracy\"], label=\"train_acc\")\nplt.plot(np.arange(0, N), history1.history[\"val_accuracy\"], label=\"val_acc\")\nplt.title(\"Training Loss and Accuracy on FER Dataset\")\nplt.xlabel(\"Epoch #\")\nplt.ylabel(\"Loss/Accuracy\")\nplt.legend(loc=\"lower left\")","execution_count":16,"outputs":[{"output_type":"execute_result","execution_count":16,"data":{"text/plain":"<matplotlib.legend.Legend at 0x7fa7501b0c10>"},"metadata":{}},{"output_type":"display_data","data":{"text/plain":"<Figure size 432x288 with 1 Axes>","image/png":"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\n"},"metadata":{}}]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.save('VGG_16_modified.h5')","execution_count":17,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.save_weights('VGG16_modified_weights.h5')","execution_count":18,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"from keras.models import model_from_json\n# serialize model to json\njson_model = model.to_json()\n#save the model architecture to JSON file\nwith open('VGG16_modified_model.json', 'w') as json_file:\n    json_file.write(json_model)","execution_count":22,"outputs":[]},{"metadata":{"trusted":true},"cell_type":"code","source":"test = '../input/fer-dsc-v3/final_data_v5/final_data_v5/Test'\ntest_datagen = ImageDataGenerator(rescale=1./255)\ntest_generator = test_datagen.flow_from_directory (test, batch_size=256, class_mode='categorical', target_size=(48,48))","execution_count":23,"outputs":[{"output_type":"stream","text":"Found 1649 images belonging to 4 classes.\n","name":"stdout"}]},{"metadata":{"trusted":true},"cell_type":"code","source":"model.evaluate(test_generator, steps=4)","execution_count":24,"outputs":[{"output_type":"stream","text":"4/4 [==============================] - 0s 90ms/step - loss: 0.7792 - accuracy: 0.7539\n","name":"stdout"},{"output_type":"execute_result","execution_count":24,"data":{"text/plain":"[0.779201865196228, 0.75390625]"},"metadata":{}}]}],"metadata":{"kernelspec":{"name":"python3","display_name":"Python 3","language":"python"},"language_info":{"name":"python","version":"3.7.6","mimetype":"text/x-python","codemirror_mode":{"name":"ipython","version":3},"pygments_lexer":"ipython3","nbconvert_exporter":"python","file_extension":".py"}},"nbformat":4,"nbformat_minor":4}