Table 3.
Architecture of expert 2
| Input image | Layer | Layer type | Size | Output shape |
|---|---|---|---|---|
| Across x axis | 1 | Convolution + ReLu | Filters 10, Kernel: 3 × 3 | (10, 93, 66) |
| Across y axis | (10, 77, 66) | |||
| Across z axis | (10, 77, 93) | |||
| Across x axis | 2 | Max pooling | Kernel 2 × 2, stride 2 | (10, 46, 33) |
| Across y axis | (10, 38, 33) | |||
| Across z axis | (10, 38, 46) | |||
| Across x axis | 3 | Convolution + ReLu | Filters 10, kernel 3 × 3 | (10, 44, 31) |
| Across y axis | (10, 36, 31) | |||
| Across z axis | (10, 36, 44) | |||
| Across x axis | 4 | Max pooling | Kernel 2 × 2, stride 2 | (10, 22, 15) |
| Across y axis | (10, 18, 15) | |||
| Across z axis | (10, 18,22) | |||
| Across x axis | 5 | Convolution + ReLu | Filters 10, kernel 3 × 3 | (10, 20, 13) |
| Across y axis | (10, 16, 13) | |||
| Across z axis | (10, 16, 20) | |||
| Across x axis | 6 | Max pooling | Kernel 2 × 2, stride 2 | (10, 10, 6) |
| Across y axis | (10, 8, 6) | |||
| Across z axis | (10, 8, 10) | |||
| Across x axis | 7 | Fully connected + ReLu dropout (rate 0.5) | 400 hidden units | 400 |
| Across y axis | 400 | |||
| Across z axis | 400 | |||
| Across x axis | 8 | Softmax | 2 | 2 |
| Across y axis | 2 | |||
| Across z axis | 2 |