Table 3 Architecture with no-separable convolution, with AVG pooling (ARCH-3)
From: Separable convolutional neural networks for facial expressions recognition
Layer (type) | Output shape | Param # |
|---|---|---|
input_1 (InputLayer) | [(None, 48, 48, 1)] | 0 |
conv2d (Conv2D) | (None, 48, 48, 8) | 72 |
batch_normalization (BN) | (None, 48, 48, 8) | 32 |
activation (Activation) | (None, 48, 48, 8) | 0 |
conv2d_1 (Conv2D) | (None, 48, 48, 8) | 576 |
batch_normalization_1 (BN) | (None, 48, 48, 8) | 32 |
activation_1 (Activation) | (None, 48, 48, 8) | 0 |
conv2d_2 (Conv2D) | (None, 48, 48, 16) | 1152 |
batch_normalization_2 (BN) | (None, 48, 48, 8) | 32 |
activation_2 (Activation) | (None, 48, 48, 16) | 0 |
conv2d_3 (Conv2D) | (None, 48, 48, 16) | 2304 |
batch_normalization_3 (BN) | (None, 48, 48, 8) | 32 |
activation_3 (Activation) | (None, 48, 48, 16) | 0 |
conv2d_4 (Conv2D) | (None, 48, 48, 16) | 2304 |
batch_normalization_4 (BN) | (None, 48, 48, 8) | 32 |
max_pooling2d (MaxPooling2D) | (None, 24, 24, 16) | 0 |
activation_4 (Activation) | (None, 24, 24, 16) | 0 |
conv2d_5 (Conv2D) | (None, 24, 24, 32) | 4608 |
batch_normalization_5 (BN) | (None, 48, 48, 8) | 32 |
activation_5 (Activation) | (None, 24, 24, 32) | 0 |
conv2d_6 (Conv2D) | (None, 24, 24, 32) | 9216 |
batch_normalization_6 (BN) | (None, 48, 48, 8) | 32 |
max_pooling2d_1 (MaxPooling2D) | (None, 12, 12, 32) | 0 |
activation_6 (Activation) | (None, 12, 12, 32) | 0 |
conv2d_7 (Conv2D) | (None, 12, 12, 64) | 18432 |
batch_normalization_7 (BN) | (None, 48, 48, 8) | 32 |
activation_7 (Activation) | (None, 12, 12, 64) | 0 |
conv2d_8 (Conv2D) | (None, 12, 12, 64) | 36864 |
batch_normalization_8 (BN) | (None, 48, 48, 8) | 32 |
max_pooling2d_2 (MaxPooling2D) | (None, 12, 12, 32) | 0 |
activation_8 (Activation) | (None, 6, 6, 64) | 0 |
conv2d_9 (Conv2D) | (None, 6, 6, 128) | 73728 |
batch_normalization_9 (BN) | (None, 48, 48, 8) | 32 |
activation_9 (Activation) | (None, 6, 6, 128) | 0 |
conv2d_10 (Conv2D) | (None, 6, 6, 128) | 147456 |
batch_normalization_10 (BN) | (None, 48, 48, 8) | 32 |
activation_10 (Activation) | (None, 6, 6, 128) | 0 |
conv2d_11 (Conv2D) | (None, 6, 6, 7) | 8071 |
global_average_pooling2d (Global AVG) | (None, 7) | 0 |
predictions (Activation) | (None, 7) | 0 |