Table 3 Architecture hyperparameters for our proposed 3D-CNN model
Layer | Filter size | Number of filters | Stride size | Dropout rate | Output size |
|---|---|---|---|---|---|
Conv1 + BN + ELU | 3 × 3 × 3 | 11 | 1 | – | 11 × 79 × 95 × 69 |
MaxPool1 | 2 × 2 × 2 | – | 2 | – | 11 × 40 × 48 × 35 |
Conv2 + BN + ELU | 3 × 3 × 3 | 11 | 1 | – | 11 × 40 × 48 × 35 |
MaxPool2 | 2 × 2 × 2 | – | 2 | – | 11 × 20 × 24 × 18 |
Conv3 + BN + ELU | 3 × 3 × 3 | 11 | 1 | – | 11 × 20 × 24 × 18 |
MaxPool3 | 2 × 2 × 2 | – | 2 | – | 11 × 10 × 12 × 9 |
Conv4 + BN + ELU | 3 × 3 × 3 | 11 | 1 | – | 11 × 10 × 12 × 9 |
MaxPool4 | 2 × 2 × 2 | – | 2 | – | 11 × 5 × 6 × 5 |
Conv5 + BN + ELU | 3 × 3 × 3 | 11 | 1 | – | 11 × 5 × 6 × 5 |
MaxPool5 | 2 × 2 × 2 | – | 2 | – | 11 × 3 × 3 × 3 |
FC 1 | – | 300 | – | – | 300 |
FC 2 | – | 100 | – | – | 100 |
FC 3 | – | 3 | – | – | 3 |
Dropout | – | – | – | 0.1 | 3 |
Softmax | – | – | – | – | 3 |