Table 8 Number of parameters for different network architectures. Among them, Inference Time (M) represents the time required for the algorithm to reason about the test set in the Synapse dataset, and M represents minutes.
From: Medical image segmentation model based on local enhancement driven global optimization
Methods | parameters(M) | FLOPs(G) | Inference Time(M) | GPU Memory(M) |
|---|---|---|---|---|
U-Net4 | 34.54 | 50.18 | 40.40 | 875 |
R50 U-Net7 | 85.75 | 37.03 | 32.63 | 1302 |
R50 Att-UNet7 | 87.14 | 37.71 | 31.33 | 1306 |
Att-UNet8 | 34.88 | 51.03 | 31.25 | 878 |
TransUNet19 | 105.32 | 24.63 | 24.57 | 1388 |
MT-UNet21 | 79.07 | 44.75 | 38.40 | 1174 |
TransClaw23 | 113.02 | 38.08 | 32.93 | 1444 |
SwinUNet32 | 27.17 | 5.92 | 32.55 | 726 |
TransDeeplab41 | 28.61 | 17.08 | 34.32 | 814 |
Ours | 25.47 | 17.46 | 23.87 | 894 |
