Table 3. Parameters of the DiatomNet architecture.
| Type | Patch size/stride | Output size | Depth | #1 × 1 | #3 × 3 reduce | #3 × 3 | #5 × 5 reduce | #5 × 5 | Pool Proj |
|---|---|---|---|---|---|---|---|---|---|
| Convolutional | 7 × 7/2 | 216 × 64 × 64 | 1 | – | – | – | – | – | – |
| Max pool | 3 × 3/2 | 108 × 32 × 64 | 0 | – | – | – | – | – | – |
| Convolutional | 3 × 3/1 | 108 × 32 × 192 | 2 | – | 64 | – | – | – | – |
| Max pool | 3 × 3/2 | 54 × 16 × 192 | 0 | – | – | – | – | – | – |
| Inception | – | 54 × 16 × 256 | 2 | 64 | 96 | 128 | 16 | 32 | 32 |
| Max pool | 3 × 3/2 | 27 × 8 × 256 | 0 | – | – | – | – | – | – |
| Inception | – | 27 × 8 × 512 | 2 | 192 | 96 | 208 | 16 | 48 | 64 |
| Max pool | 3 × 3/2 | 13 × 4 × 512 | 0 | – | – | – | – | – | – |
| Inception | – | 6 × 2 × 1,024 | 2 | 384 | 192 | 384 | 48 | 128 | 128 |
| Avg pool | 6 × 2/1 | 1 × 1 × 1,024 | 0 | – | – | – | – | – | – |
| Dropout (40%) | – | 1 × 1 × 1,024 | 0 | – | – | – | – | – | – |
| Linear | – | 1 × 1 × 68 | 1 | – | – | – | – | – | – |
| Softmax | – | 1 × 1 × 68 | 0 | – | – | – | – | – | – |