Table 1 YOLOv2 network architecture parameters for zebrafish detection.
From: Zebrafish automatic monitoring system for conditioning and behavioral analysis
Layer | Operation | Filters | Kernel size | Stride | Activations |
|---|---|---|---|---|---|
1 | Convolution—ReLu | 16 | 3 × 3 | 1 | 480 × 480 × 16 |
Batch normalization | 16 | 1 × 1 | – | 480 × 480 × 16 | |
2 | Max pooling | – | – | 2 × 2 | 240 × 240 × 32 |
Convolution—ReLu | 32 | 3 × 3 | 1 | 240 × 240 × 32 | |
Batch normalization | 32 | 1 × 1 | – | 240 × 240 × 32 | |
3 | Max pooling | – | – | 2 × 2 | 120 × 120 × 64 |
Convolution—ReLu | 64 | 3 × 3 | 1 | 120 × 120 × 64 | |
Batch normalization | 64 | 1 × 1 | – | 120 × 120 × 64 | |
4 | Max pooling | – | – | 2 × 2 | 60 × 60 × 128 |
Convolution—ReLu | 128 | 3 × 3 | 1 | 60 × 60 × 128 | |
Batch normalization | 128 | 1 × 1 | – | 60 × 60 × 128 | |
5 | Convolution—ReLu | 128 | 3 × 3 | 1 | 60 × 60 × 128 |
Batch normalization | 128 | 1 × 1 | – | 60 × 60 × 128 | |
6 | Convolution—ReLu | 128 | 3 × 3 | 1 | 60 × 60 × 128 |
Batch normalization | 128 | 1 × 1 | – | 60 × 60 × 128 | |
7 | Convolution—ReLu | 24 | 1 × 1 | 1 | 60 × 60 × 24 |
Anchor | 24 | – | – | 60 × 60 × 24 |
