Table 1 Architecture of patch-based and pixel-based models
From: Deep learning for histopathological segmentation of smooth muscle in the urinary bladder
Patch-based models | Pixel-based models | ||||||
|---|---|---|---|---|---|---|---|
Network | Depth (layers) | Trainable parameters (× 106) | Training time (hours) | Network | Encoder | Trainable parameters (× 106) | Training time (hours) |
VGG16 | 16 | 119.55 | 11.16 | U-Net | ResNet18 | 14.33 | 14.10 |
ResNet18 | 18 | 11.18 | 4.72 | MA-Net | 21.68 | 16.47 | |
SqueezeNet | 18 | 0.74 | 4.40 | DeepLabv3 +  | 12.33 | 12.58 | |
MobileNetV2 | 53 | 2.23 | 6.19 | FPN | 13.05 | 11.74 | |