EmotionDetectionCV/modelTrainer.ipynb

Load Data

# importing required modules
from zipfile import ZipFile
  
# specifying the zip file name
file_name = "/content/face-expression-recognition-dataset.zip"
  
# opening the zip file in READ mode
with ZipFile(file_name, 'r') as zip:
    # printing all the contents of the zip file
    zip.printdir()
  
    # extracting all the files
    print('Extracting all the files now...')
    zip.extractall()
    print('Done!')

import numpy as np # linear algebra
import pandas as pd # data processing, CSV file I/O (e.g. pd.read_csv)
import os
import matplotlib.pyplot as plt

from tensorflow.keras.utils import load_img, img_to_array
from keras.preprocessing.image import ImageDataGenerator
from keras.layers import Dense, Input, Dropout, GlobalAveragePooling2D
from keras.layers import Flatten
from keras.layers import Conv2D
from keras.layers import BatchNormalization
from keras.layers import Activation
from keras.layers import MaxPooling2D
from keras.models import Model, Sequential
#from tensorflow.python.keras.optimizer_v2.adam import Adam
#from keras.optimizers import adam_v2
from tensorflow.keras.optimizers import RMSprop
#from keras.optimizers import rmsprop
from tensorflow.keras.optimizers import SGD
from tensorflow.keras.optimizers import Adam
#, SGD,RMSpro

picture_size = 64
folder_path = "/content/images/"

expression = 'happy'

plt.figure(figsize= (12,12))
for i in range(1, 10, 1):
    plt.subplot(3,3,i)
    img = load_img(folder_path+"train/"+expression+"/"+
                  os.listdir(folder_path + "train/" + expression)[i], target_size=(picture_size, picture_size))
    plt.imshow(img)   
plt.show()

Image augmentation using keras ImageDataGenerator

batch_size  = 128

datagen_train  = ImageDataGenerator()
datagen_val = ImageDataGenerator()

train_set = datagen_train.flow_from_directory(folder_path+"train",
                                              target_size = (picture_size,picture_size),
                                              color_mode = "grayscale",
                                              batch_size=batch_size,
                                              class_mode='categorical',
                                              shuffle=True)


test_set = datagen_val.flow_from_directory(folder_path+"validation",
                                              target_size = (picture_size,picture_size),
                                              color_mode = "grayscale",
                                              batch_size=batch_size,
                                              class_mode='categorical',
                                              shuffle=False)

Found 28821 images belonging to 7 classes.
Found 7066 images belonging to 7 classes.

Define model

no_of_classes = 7

model = Sequential()

#1st CNN layer
model.add(Conv2D(64,(3,3),padding = 'same',input_shape = (picture_size,picture_size,1)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2)))
model.add(Dropout(0.25))

#2nd CNN layer
model.add(Conv2D(128,(5,5),padding = 'same'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2)))
model.add(Dropout (0.25))

#3rd CNN layer
model.add(Conv2D(512,(3,3),padding = 'same'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size = (2,2)))
model.add(Dropout (0.25))

#4th CNN layer
model.add(Conv2D(512,(3,3), padding='same'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2, 2)))
model.add(Dropout(0.25))

model.add(Flatten())

#Fully connected 1st layer
model.add(Dense(256))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dropout(0.25))


# Fully connected layer 2nd layer
model.add(Dense(512))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dropout(0.25))

model.add(Dense(no_of_classes, activation='softmax'))

model.summary()

Model: "sequential_9"
_________________________________________________________________
 Layer (type)                Output Shape              Param #   
=================================================================
 conv2d_36 (Conv2D)          (None, 64, 64, 64)        640       
                                                                 
 batch_normalization_54 (Bat  (None, 64, 64, 64)       256       
 chNormalization)                                                
                                                                 
 activation_54 (Activation)  (None, 64, 64, 64)        0         
                                                                 
 max_pooling2d_36 (MaxPoolin  (None, 32, 32, 64)       0         
 g2D)                                                            
                                                                 
 dropout_54 (Dropout)        (None, 32, 32, 64)        0         
                                                                 
 conv2d_37 (Conv2D)          (None, 32, 32, 128)       204928    
                                                                 
 batch_normalization_55 (Bat  (None, 32, 32, 128)      512       
 chNormalization)                                                
                                                                 
 activation_55 (Activation)  (None, 32, 32, 128)       0         
                                                                 
 max_pooling2d_37 (MaxPoolin  (None, 16, 16, 128)      0         
 g2D)                                                            
                                                                 
 dropout_55 (Dropout)        (None, 16, 16, 128)       0         
                                                                 
 conv2d_38 (Conv2D)          (None, 16, 16, 512)       590336    
                                                                 
 batch_normalization_56 (Bat  (None, 16, 16, 512)      2048      
 chNormalization)                                                
                                                                 
 activation_56 (Activation)  (None, 16, 16, 512)       0         
                                                                 
 max_pooling2d_38 (MaxPoolin  (None, 8, 8, 512)        0         
 g2D)                                                            
                                                                 
 dropout_56 (Dropout)        (None, 8, 8, 512)         0         
                                                                 
 conv2d_39 (Conv2D)          (None, 8, 8, 512)         2359808   
                                                                 
 batch_normalization_57 (Bat  (None, 8, 8, 512)        2048      
 chNormalization)                                                
                                                                 
 activation_57 (Activation)  (None, 8, 8, 512)         0         
                                                                 
 max_pooling2d_39 (MaxPoolin  (None, 4, 4, 512)        0         
 g2D)                                                            
                                                                 
 dropout_57 (Dropout)        (None, 4, 4, 512)         0         
                                                                 
 flatten_9 (Flatten)         (None, 8192)              0         
                                                                 
 dense_27 (Dense)            (None, 256)               2097408   
                                                                 
 batch_normalization_58 (Bat  (None, 256)              1024      
 chNormalization)                                                
                                                                 
 activation_58 (Activation)  (None, 256)               0         
                                                                 
 dropout_58 (Dropout)        (None, 256)               0         
                                                                 
 dense_28 (Dense)            (None, 512)               131584    
                                                                 
 batch_normalization_59 (Bat  (None, 512)              2048      
 chNormalization)                                                
                                                                 
 activation_59 (Activation)  (None, 512)               0         
                                                                 
 dropout_59 (Dropout)        (None, 512)               0         
                                                                 
 dense_29 (Dense)            (None, 7)                 3591      
                                                                 
=================================================================
Total params: 5,396,231
Trainable params: 5,392,263
Non-trainable params: 3,968
_________________________________________________________________

from keras.callbacks import ModelCheckpoint, EarlyStopping, ReduceLROnPlateau
import tensorflow as tf

checkpoint = ModelCheckpoint("./model_weights.h5", monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')

reduce_learningrate = ReduceLROnPlateau(monitor='val_loss',
                              factor=0.2,
                              patience=3,
                              verbose=1,
                              min_delta=0.0001)

early = EarlyStopping(monitor='val_accuracy', min_delta=0, patience=5, verbose=1, mode='auto')

callbacks_list = [checkpoint,reduce_learningrate, early]

epochs = 100

model.compile(loss='categorical_crossentropy',
              optimizer = Adam(lr=0.001),
              metrics=['accuracy', tf.keras.metrics.Precision(), tf.keras.metrics.Recall()])

Train model

history = model.fit(train_set,
                    steps_per_epoch=train_set.n//train_set.batch_size,
                    epochs=epochs,
                    validation_data = test_set,
                    validation_steps = test_set.n//test_set.batch_size,
                    callbacks=callbacks_list
                    )

Epoch 1/100
225/225 [==============================] - ETA: 0s - loss: 1.7420 - accuracy: 0.3358 - precision: 0.5071 - recall: 0.1418
Epoch 1: val_accuracy improved from -inf to 0.41293, saving model to ./model_weights.h5
225/225 [==============================] - 31s 90ms/step - loss: 1.7420 - accuracy: 0.3358 - precision: 0.5071 - recall: 0.1418 - val_loss: 1.5676 - val_accuracy: 0.4129 - val_precision: 0.5814 - val_recall: 0.2881 - lr: 0.0010
Epoch 2/100
225/225 [==============================] - ETA: 0s - loss: 1.4073 - accuracy: 0.4584 - precision: 0.6749 - recall: 0.2597
Epoch 2: val_accuracy improved from 0.41293 to 0.46293, saving model to ./model_weights.h5
225/225 [==============================] - 20s 88ms/step - loss: 1.4073 - accuracy: 0.4584 - precision: 0.6749 - recall: 0.2597 - val_loss: 1.5090 - val_accuracy: 0.4629 - val_precision: 0.5566 - val_recall: 0.3661 - lr: 0.0010
Epoch 3/100
225/225 [==============================] - ETA: 0s - loss: 1.2546 - accuracy: 0.5188 - precision: 0.7175 - recall: 0.3240
Epoch 3: val_accuracy improved from 0.46293 to 0.52088, saving model to ./model_weights.h5
225/225 [==============================] - 20s 89ms/step - loss: 1.2546 - accuracy: 0.5188 - precision: 0.7175 - recall: 0.3240 - val_loss: 1.2621 - val_accuracy: 0.5209 - val_precision: 0.7545 - val_recall: 0.2872 - lr: 0.0010
Epoch 4/100
225/225 [==============================] - ETA: 0s - loss: 1.1689 - accuracy: 0.5555 - precision: 0.7336 - recall: 0.3686
Epoch 4: val_accuracy did not improve from 0.52088
225/225 [==============================] - 19s 86ms/step - loss: 1.1689 - accuracy: 0.5555 - precision: 0.7336 - recall: 0.3686 - val_loss: 1.2827 - val_accuracy: 0.5192 - val_precision: 0.7136 - val_recall: 0.3149 - lr: 0.0010
Epoch 5/100
225/225 [==============================] - ETA: 0s - loss: 1.1016 - accuracy: 0.5805 - precision: 0.7429 - recall: 0.4076
Epoch 5: val_accuracy improved from 0.52088 to 0.57514, saving model to ./model_weights.h5
225/225 [==============================] - 20s 87ms/step - loss: 1.1016 - accuracy: 0.5805 - precision: 0.7429 - recall: 0.4076 - val_loss: 1.1303 - val_accuracy: 0.5751 - val_precision: 0.7255 - val_recall: 0.4307 - lr: 0.0010
Epoch 6/100
225/225 [==============================] - ETA: 0s - loss: 1.0435 - accuracy: 0.6037 - precision: 0.7507 - recall: 0.4417
Epoch 6: val_accuracy did not improve from 0.57514
225/225 [==============================] - 19s 86ms/step - loss: 1.0435 - accuracy: 0.6037 - precision: 0.7507 - recall: 0.4417 - val_loss: 1.2345 - val_accuracy: 0.5409 - val_precision: 0.6644 - val_recall: 0.4257 - lr: 0.0010
Epoch 7/100
225/225 [==============================] - ETA: 0s - loss: 0.9941 - accuracy: 0.6260 - precision: 0.7590 - recall: 0.4753
Epoch 7: val_accuracy did not improve from 0.57514
225/225 [==============================] - 20s 87ms/step - loss: 0.9941 - accuracy: 0.6260 - precision: 0.7590 - recall: 0.4753 - val_loss: 1.2263 - val_accuracy: 0.5273 - val_precision: 0.6728 - val_recall: 0.3714 - lr: 0.0010
Epoch 8/100
225/225 [==============================] - ETA: 0s - loss: 0.9451 - accuracy: 0.6442 - precision: 0.7697 - recall: 0.5077
Epoch 8: val_accuracy did not improve from 0.57514

Epoch 8: ReduceLROnPlateau reducing learning rate to 0.00020000000949949026.
225/225 [==============================] - 20s 87ms/step - loss: 0.9451 - accuracy: 0.6442 - precision: 0.7697 - recall: 0.5077 - val_loss: 1.1971 - val_accuracy: 0.5574 - val_precision: 0.6877 - val_recall: 0.4463 - lr: 0.0010
Epoch 9/100
225/225 [==============================] - ETA: 0s - loss: 0.8071 - accuracy: 0.7006 - precision: 0.8094 - recall: 0.5765
Epoch 9: val_accuracy improved from 0.57514 to 0.62713, saving model to ./model_weights.h5
225/225 [==============================] - 20s 88ms/step - loss: 0.8071 - accuracy: 0.7006 - precision: 0.8094 - recall: 0.5765 - val_loss: 1.0264 - val_accuracy: 0.6271 - val_precision: 0.7245 - val_recall: 0.5330 - lr: 2.0000e-04
Epoch 10/100
225/225 [==============================] - ETA: 0s - loss: 0.7597 - accuracy: 0.7165 - precision: 0.8103 - recall: 0.6141
Epoch 10: val_accuracy improved from 0.62713 to 0.63778, saving model to ./model_weights.h5
225/225 [==============================] - 20s 87ms/step - loss: 0.7597 - accuracy: 0.7165 - precision: 0.8103 - recall: 0.6141 - val_loss: 1.0111 - val_accuracy: 0.6378 - val_precision: 0.7350 - val_recall: 0.5436 - lr: 2.0000e-04
Epoch 11/100
225/225 [==============================] - ETA: 0s - loss: 0.7127 - accuracy: 0.7345 - precision: 0.8185 - recall: 0.6381
Epoch 11: val_accuracy did not improve from 0.63778
225/225 [==============================] - 19s 87ms/step - loss: 0.7127 - accuracy: 0.7345 - precision: 0.8185 - recall: 0.6381 - val_loss: 1.0742 - val_accuracy: 0.6089 - val_precision: 0.6938 - val_recall: 0.5284 - lr: 2.0000e-04
Epoch 12/100
225/225 [==============================] - ETA: 0s - loss: 0.6812 - accuracy: 0.7474 - precision: 0.8247 - recall: 0.6629
Epoch 12: val_accuracy did not improve from 0.63778
225/225 [==============================] - 20s 87ms/step - loss: 0.6812 - accuracy: 0.7474 - precision: 0.8247 - recall: 0.6629 - val_loss: 1.0393 - val_accuracy: 0.6374 - val_precision: 0.7161 - val_recall: 0.5658 - lr: 2.0000e-04
Epoch 13/100
225/225 [==============================] - ETA: 0s - loss: 0.6441 - accuracy: 0.7623 - precision: 0.8317 - recall: 0.6828
Epoch 13: val_accuracy did not improve from 0.63778

Epoch 13: ReduceLROnPlateau reducing learning rate to 4.0000001899898055e-05.
225/225 [==============================] - 19s 86ms/step - loss: 0.6441 - accuracy: 0.7623 - precision: 0.8317 - recall: 0.6828 - val_loss: 1.0473 - val_accuracy: 0.6325 - val_precision: 0.7094 - val_recall: 0.5625 - lr: 2.0000e-04
Epoch 14/100
225/225 [==============================] - ETA: 0s - loss: 0.5993 - accuracy: 0.7806 - precision: 0.8453 - recall: 0.7078
Epoch 14: val_accuracy improved from 0.63778 to 0.64986, saving model to ./model_weights.h5
225/225 [==============================] - 20s 87ms/step - loss: 0.5993 - accuracy: 0.7806 - precision: 0.8453 - recall: 0.7078 - val_loss: 1.0256 - val_accuracy: 0.6499 - val_precision: 0.7211 - val_recall: 0.5791 - lr: 4.0000e-05
Epoch 15/100
225/225 [==============================] - ETA: 0s - loss: 0.5814 - accuracy: 0.7874 - precision: 0.8508 - recall: 0.7183
Epoch 15: val_accuracy did not improve from 0.64986
225/225 [==============================] - 19s 87ms/step - loss: 0.5814 - accuracy: 0.7874 - precision: 0.8508 - recall: 0.7183 - val_loss: 1.0313 - val_accuracy: 0.6473 - val_precision: 0.7133 - val_recall: 0.5827 - lr: 4.0000e-05
Epoch 16/100
225/225 [==============================] - ETA: 0s - loss: 0.5639 - accuracy: 0.7922 - precision: 0.8535 - recall: 0.7272
Epoch 16: val_accuracy improved from 0.64986 to 0.65142, saving model to ./model_weights.h5

Epoch 16: ReduceLROnPlateau reducing learning rate to 8.000000525498762e-06.
225/225 [==============================] - 20s 87ms/step - loss: 0.5639 - accuracy: 0.7922 - precision: 0.8535 - recall: 0.7272 - val_loss: 1.0401 - val_accuracy: 0.6514 - val_precision: 0.7128 - val_recall: 0.5865 - lr: 4.0000e-05
Epoch 17/100
225/225 [==============================] - ETA: 0s - loss: 0.5567 - accuracy: 0.7941 - precision: 0.8535 - recall: 0.7311
Epoch 17: val_accuracy did not improve from 0.65142
225/225 [==============================] - 20s 87ms/step - loss: 0.5567 - accuracy: 0.7941 - precision: 0.8535 - recall: 0.7311 - val_loss: 1.0323 - val_accuracy: 0.6514 - val_precision: 0.7176 - val_recall: 0.5898 - lr: 8.0000e-06
Epoch 18/100
225/225 [==============================] - ETA: 0s - loss: 0.5549 - accuracy: 0.7951 - precision: 0.8538 - recall: 0.7285
Epoch 18: val_accuracy improved from 0.65142 to 0.65213, saving model to ./model_weights.h5
225/225 [==============================] - 20s 87ms/step - loss: 0.5549 - accuracy: 0.7951 - precision: 0.8538 - recall: 0.7285 - val_loss: 1.0329 - val_accuracy: 0.6521 - val_precision: 0.7159 - val_recall: 0.5892 - lr: 8.0000e-06
Epoch 19/100
225/225 [==============================] - ETA: 0s - loss: 0.5445 - accuracy: 0.7993 - precision: 0.8575 - recall: 0.7373
Epoch 19: val_accuracy improved from 0.65213 to 0.65256, saving model to ./model_weights.h5

Epoch 19: ReduceLROnPlateau reducing learning rate to 1.6000001778593287e-06.
225/225 [==============================] - 20s 87ms/step - loss: 0.5445 - accuracy: 0.7993 - precision: 0.8575 - recall: 0.7373 - val_loss: 1.0337 - val_accuracy: 0.6526 - val_precision: 0.7167 - val_recall: 0.5899 - lr: 8.0000e-06
Epoch 20/100
225/225 [==============================] - ETA: 0s - loss: 0.5475 - accuracy: 0.7987 - precision: 0.8584 - recall: 0.7367
Epoch 20: val_accuracy improved from 0.65256 to 0.65270, saving model to ./model_weights.h5
225/225 [==============================] - 20s 87ms/step - loss: 0.5475 - accuracy: 0.7987 - precision: 0.8584 - recall: 0.7367 - val_loss: 1.0340 - val_accuracy: 0.6527 - val_precision: 0.7180 - val_recall: 0.5905 - lr: 1.6000e-06
Epoch 21/100
225/225 [==============================] - ETA: 0s - loss: 0.5481 - accuracy: 0.7980 - precision: 0.8559 - recall: 0.7342
Epoch 21: val_accuracy did not improve from 0.65270
225/225 [==============================] - 20s 87ms/step - loss: 0.5481 - accuracy: 0.7980 - precision: 0.8559 - recall: 0.7342 - val_loss: 1.0343 - val_accuracy: 0.6524 - val_precision: 0.7179 - val_recall: 0.5906 - lr: 1.6000e-06
Epoch 22/100
225/225 [==============================] - ETA: 0s - loss: 0.5465 - accuracy: 0.8001 - precision: 0.8579 - recall: 0.7357
Epoch 22: val_accuracy did not improve from 0.65270

Epoch 22: ReduceLROnPlateau reducing learning rate to 3.200000264769187e-07.
225/225 [==============================] - 19s 87ms/step - loss: 0.5465 - accuracy: 0.8001 - precision: 0.8579 - recall: 0.7357 - val_loss: 1.0348 - val_accuracy: 0.6518 - val_precision: 0.7165 - val_recall: 0.5911 - lr: 1.6000e-06
Epoch 23/100
225/225 [==============================] - ETA: 0s - loss: 0.5509 - accuracy: 0.7977 - precision: 0.8551 - recall: 0.7329
Epoch 23: val_accuracy improved from 0.65270 to 0.65298, saving model to ./model_weights.h5
225/225 [==============================] - 20s 88ms/step - loss: 0.5509 - accuracy: 0.7977 - precision: 0.8551 - recall: 0.7329 - val_loss: 1.0352 - val_accuracy: 0.6530 - val_precision: 0.7173 - val_recall: 0.5906 - lr: 3.2000e-07
Epoch 24/100
225/225 [==============================] - ETA: 0s - loss: 0.5495 - accuracy: 0.7991 - precision: 0.8578 - recall: 0.7371
Epoch 24: val_accuracy did not improve from 0.65298
225/225 [==============================] - 19s 86ms/step - loss: 0.5495 - accuracy: 0.7991 - precision: 0.8578 - recall: 0.7371 - val_loss: 1.0349 - val_accuracy: 0.6530 - val_precision: 0.7172 - val_recall: 0.5909 - lr: 3.2000e-07
Epoch 25/100
225/225 [==============================] - ETA: 0s - loss: 0.5462 - accuracy: 0.7996 - precision: 0.8584 - recall: 0.7371
Epoch 25: val_accuracy improved from 0.65298 to 0.65312, saving model to ./model_weights.h5

Epoch 25: ReduceLROnPlateau reducing learning rate to 6.400000529538374e-08.
225/225 [==============================] - 20s 87ms/step - loss: 0.5462 - accuracy: 0.7996 - precision: 0.8584 - recall: 0.7371 - val_loss: 1.0344 - val_accuracy: 0.6531 - val_precision: 0.7175 - val_recall: 0.5903 - lr: 3.2000e-07
Epoch 26/100
225/225 [==============================] - ETA: 0s - loss: 0.5404 - accuracy: 0.8024 - precision: 0.8617 - recall: 0.7379
Epoch 26: val_accuracy did not improve from 0.65312
225/225 [==============================] - 19s 86ms/step - loss: 0.5404 - accuracy: 0.8024 - precision: 0.8617 - recall: 0.7379 - val_loss: 1.0357 - val_accuracy: 0.6524 - val_precision: 0.7160 - val_recall: 0.5906 - lr: 6.4000e-08
Epoch 27/100
225/225 [==============================] - ETA: 0s - loss: 0.5482 - accuracy: 0.7965 - precision: 0.8563 - recall: 0.7341
Epoch 27: val_accuracy did not improve from 0.65312
225/225 [==============================] - 19s 86ms/step - loss: 0.5482 - accuracy: 0.7965 - precision: 0.8563 - recall: 0.7341 - val_loss: 1.0346 - val_accuracy: 0.6531 - val_precision: 0.7173 - val_recall: 0.5905 - lr: 6.4000e-08
Epoch 28/100
225/225 [==============================] - ETA: 0s - loss: 0.5436 - accuracy: 0.8013 - precision: 0.8621 - recall: 0.7372
Epoch 28: val_accuracy improved from 0.65312 to 0.65327, saving model to ./model_weights.h5

Epoch 28: ReduceLROnPlateau reducing learning rate to 1.2800001059076749e-08.
225/225 [==============================] - 20s 87ms/step - loss: 0.5436 - accuracy: 0.8013 - precision: 0.8621 - recall: 0.7372 - val_loss: 1.0353 - val_accuracy: 0.6533 - val_precision: 0.7171 - val_recall: 0.5908 - lr: 6.4000e-08
Epoch 29/100
225/225 [==============================] - ETA: 0s - loss: 0.5509 - accuracy: 0.7975 - precision: 0.8563 - recall: 0.7354
Epoch 29: val_accuracy did not improve from 0.65327
225/225 [==============================] - 20s 88ms/step - loss: 0.5509 - accuracy: 0.7975 - precision: 0.8563 - recall: 0.7354 - val_loss: 1.0347 - val_accuracy: 0.6527 - val_precision: 0.7175 - val_recall: 0.5919 - lr: 1.2800e-08
Epoch 30/100
225/225 [==============================] - ETA: 0s - loss: 0.5535 - accuracy: 0.7965 - precision: 0.8578 - recall: 0.7320
Epoch 30: val_accuracy did not improve from 0.65327
225/225 [==============================] - 19s 86ms/step - loss: 0.5535 - accuracy: 0.7965 - precision: 0.8578 - recall: 0.7320 - val_loss: 1.0349 - val_accuracy: 0.6526 - val_precision: 0.7172 - val_recall: 0.5915 - lr: 1.2800e-08
Epoch 31/100
225/225 [==============================] - ETA: 0s - loss: 0.5521 - accuracy: 0.7971 - precision: 0.8559 - recall: 0.7335
Epoch 31: val_accuracy did not improve from 0.65327

Epoch 31: ReduceLROnPlateau reducing learning rate to 2.5600002118153498e-09.
225/225 [==============================] - 19s 87ms/step - loss: 0.5521 - accuracy: 0.7971 - precision: 0.8559 - recall: 0.7335 - val_loss: 1.0362 - val_accuracy: 0.6528 - val_precision: 0.7168 - val_recall: 0.5922 - lr: 1.2800e-08
Epoch 32/100
225/225 [==============================] - ETA: 0s - loss: 0.5469 - accuracy: 0.7996 - precision: 0.8587 - recall: 0.7378
Epoch 32: val_accuracy did not improve from 0.65327
225/225 [==============================] - 19s 86ms/step - loss: 0.5469 - accuracy: 0.7996 - precision: 0.8587 - recall: 0.7378 - val_loss: 1.0347 - val_accuracy: 0.6527 - val_precision: 0.7173 - val_recall: 0.5912 - lr: 2.5600e-09
Epoch 33/100
225/225 [==============================] - ETA: 0s - loss: 0.5545 - accuracy: 0.7952 - precision: 0.8534 - recall: 0.7318
Epoch 33: val_accuracy did not improve from 0.65327
225/225 [==============================] - 19s 86ms/step - loss: 0.5545 - accuracy: 0.7952 - precision: 0.8534 - recall: 0.7318 - val_loss: 1.0355 - val_accuracy: 0.6524 - val_precision: 0.7167 - val_recall: 0.5915 - lr: 2.5600e-09
Epoch 33: early stopping

plt.figure(figsize=(20,10))
plt.subplot(1, 2, 1)
plt.suptitle('Optimizer : Adam', fontsize=10)
plt.ylabel('Loss', fontsize=16)
plt.plot(history.history['loss'], label='Training Loss')
plt.plot(history.history['val_loss'], label='Validation Loss')
plt.legend(loc='upper right')

plt.subplot(1, 2, 2)
plt.ylabel('Accuracy', fontsize=16)
plt.plot(history.history['accuracy'], label='Training Accuracy')
plt.plot(history.history['val_accuracy'], label='Validation Accuracy')
plt.legend(loc='lower right')
plt.show()

plt.figure(figsize=(20,10))
plt.subplot(1, 2, 1)
plt.suptitle('Optimizer : Adam', fontsize=10)
plt.ylabel('Recall', fontsize=16)
plt.plot(history.history['recall'], label='Training Recall')
plt.plot(history.history['val_recall'], label='Validation Recall')
plt.legend(loc='upper right')

plt.subplot(1, 2, 2)
plt.ylabel('Precision', fontsize=16)
plt.plot(history.history['precision'], label='Training Precision')
plt.plot(history.history['val_precision'], label='Validation Precision')
plt.legend(loc='lower right')
plt.show()

# serialize model to JSON
model_json = model.to_json()
with open("/content/model.json", "w") as json_file:
    json_file.write(model_json)
# serialize weights to HDF5
# model.save_weights("model.h5")
print("Saved model to disk")

Saved model to disk

from tensorflow.keras.models import model_from_json

model_json_file = '/content/model.json'
model_weights_file = '/content/model_weights.h5'
with open(model_json_file, "r") as json_file:
    loaded_model_json = json_file.read()
    loaded_model = model_from_json(loaded_model_json)
    loaded_model.load_weights(model_weights_file)

images_predict = {"angry": "8969.jpg", "disgust": "14954.jpg", "fear": "10409.jpg", "happy": "10019.jpg", "neutral": "10033.jpg", "sad": "10004.jpg", "surprise": "1033.jpg"}

import cv2

Test predictions

images_path = "/content/images/images/validation/"

label_map = (train_set.class_indices)
print(label_map)

for key in images_predict:
  image_sample = cv2.imread(images_path + key + "/" + images_predict[key])
  gray = cv2.cvtColor(image_sample, cv2.COLOR_BGR2GRAY)
  roi = cv2.resize(gray, (picture_size,picture_size))
  y_probs = loaded_model.predict(roi[np.newaxis, :, :, np.newaxis])
  val  = np.argmax(y_probs, axis=-1)[0]
  for k, v in label_map.items():
    if val == v:
        print("Predicted: {}, Should be: {}".format(k,key))

{'angry': 0, 'disgust': 1, 'fear': 2, 'happy': 3, 'neutral': 4, 'sad': 5, 'surprise': 6}
1/1 [==============================] - 0s 136ms/step
Predicted: fear, Should be: angry
1/1 [==============================] - 0s 17ms/step
Predicted: neutral, Should be: disgust
1/1 [==============================] - 0s 20ms/step
Predicted: happy, Should be: fear
1/1 [==============================] - 0s 15ms/step
Predicted: happy, Should be: happy
1/1 [==============================] - 0s 14ms/step
Predicted: neutral, Should be: neutral
1/1 [==============================] - 0s 14ms/step
Predicted: angry, Should be: sad
1/1 [==============================] - 0s 19ms/step
Predicted: surprise, Should be: surprise

source