Table.3 Detailed network configuration of U-Net GAN discriminator and U-Net GAN + SA discriminator
From: A method for improving semantic segmentation using thermographic images in infants
Layers | Output size | U-Net | U-Net GAN + SA |
|---|---|---|---|
Input | 320 × 256 × 1 |  |  |
Convolution | 320 × 256 × 8 | 3 × 3, 8 d | 3 × 3, 8 d |
Downscale | 160 × 128 × 16 | 5 × 5, 16 d, CBR 3 × 3, 16 d, CBR | 1 × 1, 16 d 7 × 7, 16 d, SA 1 × 1, 16 d |
Downscale | 80 × 64 × 32 | 5 × 5, 32 d, CBR 3 × 3, 32 d, CBR | 1 × 1, 32 d 77, 32 d, SA 1 × 1, 32 d |
Downscale | 40 × 32 × 64 | 5 × 5, 64 d, CBR 3 × 3, 64 d, CBR | 1 × 1, 64 d 7 × 7, 64 d, SA 1 × 1, 64 d |
Encoder out (\(D_{enc} \left( x \right)\)) | 5 | ReLU Average Pooling Linear, 5d | ReLU Average Pooling Linear, 5 d |
Upscale | 80 × 64 × 32 | 5 × 5, 32 d, CBR 3 × 3, 32 d, CBR | 1 × 1, 32 d 7 × 7, 32 d, SA 1 × 1, 32 d |
Upscale | 160 × 128 × 16 | 5 × 5, 16 d, CBR 3 × 3, 16 d, CBR | 1 × 1, 16 d 7 × 7, 16 d, SA 1 × 1, 16 d |
Upscale | 0 × 256 × 8 | 5 × 5, 8 d, CBR 33, 8 d, CBR | 1 × 1, 8 d 7 × 7, 8 d, SA 1 × 1, 8 d |
Convolution (\(D_{dec} \left( x \right)\)) | 320 × 256 × 2 | 3 × 3, 2 d | 3 × 3, 2 d |