TABLE 1.
The Architecture of Our Proposed CNN
| Layer | #channels | #neurons | Kernel | stride |
|---|---|---|---|---|
| Convolutional | 64 | 32 × 32 | 3 × 3 | 2 |
| Convolutional | 64 | 32 × 32 | 3 × 3 | 1 |
| Max-pooling | 64 | 16 × 16 | 2 × 2 | 2 |
|
| ||||
| Convolutional | 64 | 16 × 16 | 3 × 3 | 1 |
| Convolutional | 64 | 16 × 16 | 3 × 3 | 1 |
| Max-pooling | 64 | 8 × 8 | 2 × 2 | 2 |
|
| ||||
| Convolutional | 64 | 8 × 8 | 3 × 3 | 1 |
| Convolutional | 64 | 8 × 8 | 3 × 3 | 1 |
| Max-pooling | 64 | 4 × 4 | 2 × 2 | 2 |
|
| ||||
| Fully-connected | 200 | 1 × 1 | ||
| Fully-connected | #classes | 1 × 1 | ||