Table 2.
CNN model architecture details.
| Layer | Dimension | Stride | Activation |
|---|---|---|---|
| Input | 96 × 96 × 3 | — | — |
| Convolutional layer | 30@8 × 8 | 2 | ReLU |
| Max pooling | 3 × 3 | 1 | — |
| Convolutional layer | 100@5 × 5 | 2 | ReLU |
| Max pooling | 3 × 3 | 1 | — |
| Convolutional layer | 100@5 × 5 | 2 | ReLU |
| Max pooling | 3 × 3 | 1 | — |
| Fully connected | 300 | — | ReLU |
| Fully connected | 200 | — | ReLU |
| Output | 1 | — | Sigmoid |