Table 1 The proposed CNN configuration for an input patch of size 250 × 250.
From: A pyramidal deep learning pipeline for kidney whole-slide histology images classification
Layer | Depth | kernel | Stride | Spatial Size | Parameters |
|---|---|---|---|---|---|
Input | 3 | – | – | 250 × 250 × 3 | 0 |
1. Conv | 9 | 3 × 3 | 1 × 1 | 248 × 248 × 9 | 3 × 3 × 3 × 9 |
2. Conv | 9 | 3 × 3 | 1 × 1 | 246 × 246 × 9 | 3 × 3 × 9 × 9 |
3. Max-pool | 9 | 2 × 2 | 2 × 2 | 123 × 123 × 9 | 0 |
4. Conv | 9 | 3 × 3 | 1 × 1 | 121 × 121 × 9 | 3 × 3 × 9 × 9 |
5. Conv | 9 | 3 × 3 | 1 × 1 | 119 × 119 × 9 | 3 × 3 × 9 × 9 |
6. Max-pool | 9 | 2 × 2 | 2 × 2 | 59 × 59 × 9 | 0 |
7. Conv | 9 | 3 × 3 | 1 × 1 | 57 × 57 × 9 | 3 × 3 × 9 × 9 |
8. Conv | 9 | 3 × 3 | 1 × 1 | 55 × 55 × 9 | 3 × 3 × 9 × 9 |
9. Max-pool | 9 | 2 × 2 | 2 × 2 | 27 × 27 × 9 | 0 |
10. Conv | 9 | 3 × 3 | 1 × 1 | 25 × 25 × 9 | 3 × 3 × 9 × 9 |
11. Conv | 9 | 3 × 3 | 1 × 1 | 23 × 23 × 9 | 3 × 3 × 9 × 9 |
12. Max-pool | 9 | 2 × 2 | 2 × 2 | 11 × 11 × 9 | 0 |
13. Concat | 1 | – | – | 1089 × 1 | 0 |
14. Full | 1 | – | – | 12 × 1 | 1089 × 12 |
15. Full | 1 | – | – | 4 × 1 | 12 × 4 |
16. Softmax | 1 | – | – | 4 × 1 | 0 |
