Table 3.
CNN layers for emotion recognition.
| Layer (Type) | Output Shape | Parameter Numbers | Activation |
|---|---|---|---|
| conv2d (Conv2D) | (None, 48, 48, 32) | 320 | ReLU |
| conv2d_1 (Conv2D) | (None, 48, 48, 32) | 9248 | ReLU |
| max_pooling2d (MaxPooling2D) | (None, 24, 24, 32) | 0 | ReLU |
| conv2d_2 (Conv2D) | (None, 24, 24, 64) | 18,496 | ReLU |
| conv2d_3 (Conv2D) | (None, 24, 24, 64) | 36,928 | ReLU |
| max_pooling2d_1 (MaxPooling 2D) | (None, 12, 12, 64) | 0 | ReLU |
| conv2d_4 (Conv2D) | (None, 12, 12, 128) | 73,856 | ReLU |
| conv2d_5 (Conv2D) | (None, 12, 12, 128) | 147,584 | ReLU |
| max_pooling2d_1 (MaxPooling 2D) | (None, 6, 6, 128) | 0 | ReLU |
| flatten (Flatten) | (None, 4608) | 0 | None |
| dense (Dense) | (None, 512) | 2,359,808 | ReLU |
| dense_1 (Dense) | (None, 64) | 32,832 | ReLU |
| dense_2 (Dense) | (None, 7) | 455 | Softmax |