Mastering Computer Vision with TensorFlow 2.x

Iteration 3 – CNN parameter adjustment

The second convolution layer was dropped completely; the input layer was changed from 5 x 5 to 3 x 3:

model.add(Conv2D(64, (3, 3), activation='relu', input_shape=(48,48,1)))

The third convolution layer remains unchanged. This layer is as follows:

model.add(Conv2D(128, (3, 3), activation='relu'))model.add(AveragePooling2D(pool_size=(3,3), strides=(2, 2)))

There is no change in the dense layer. The output of this is as follows:

Epoch 1/5 256/256 [==========================] - 410s 2s/step - loss: 1.6465 - acc: 0.3500 Epoch 2/5 256/256 [==========================] - 415s 2s/step - loss: 1.3435 - acc: 0.4851 Epoch 3/5 256/256 [==========================] - 412s 2s/step - loss: 1.0837 - acc: 0.5938 Epoch ...