Table 3.
Structure of CNN.
| Layer name | Filters | Kernel size | Output tensor |
|---|---|---|---|
| Convolutional layer 1 | 16 | 3×3 | 16×32×32 |
| Batch normalization 1 | 16×32×32 | ||
| Pooling layer 1 | 16×16×16 | ||
| Convolutional layer 2 | 32 | 3×3 | 32×16×16 |
| Batch normalization 2 | 32×16×16 | ||
| Pooling layer 2 | 32×8×8 | ||
| Convolutional layer 3 | 64 | 3×3 | 64×8×8 |
| Batch normalization 3 | 64×8×8 | ||
| Pooling layer 3 | 64×4×4 | ||
| Convolutional layer 4 | 128 | 3×3 | 128×4×4 |
| Batch normalization 4 | 128×4×4 | ||
| Pooling layer 4 | 128×2×2 | ||
| Dropout 1 | 128×2×2 | ||
| Flatten layer 1 | 512 | ||
| Dense layer 1 | 64 | ||
| Dense layer 2 | 1 |