From: Ancient mural classification methods based on a multichannel separable network
Layer type | Patch size/stride | Output size |
---|---|---|
Convolutional | 7 × 7/2 | 112 × 112 × 64 |
Max pooling | 3 × 3/2 | 56 × 56 × 64 |
Convolutional | 3 × 3/1 | 56 × 56 × 192 |
Max pooling | 3 × 3/2 | 28 × 28 × 192 |
Inception (3a) | 28 × 28 × 256 | |
Inception (3b) | 28 × 28 × 480 | |
Max pooling | 3 × 3/2 | 14 × 14 × 480 |
Inception (4a) | 14 × 14 × 512 | |
Inception (4b) | 14 × 14 × 512 | |
Inception (4c) | 14 × 14 × 512 | |
Inception (4d) | 14 × 14 × 528 | |
Inception (4e) | 14 × 14 × 832 | |
Max pooling | 3 × 3/2 | 7 × 7 × 832 |
Inception (5a) | 7 × 7 × 832 | |
Inception (5b) | 7 × 7 × 1024 | |
Avg pooling | 7 × 7/2 | 1 × 1 × 1024 |
Dropout (40%) | 1 × 1 × 1024 | |
Softmax with loss |