Comparative performance results on models’ computations and the number of parameters
| Models | Year | Params (millions)↓ | GFLOPS↓ |
|---|---|---|---|
| U-Net | 2015 | 2.009 | 3.224 |
| SCR-Net | 2021 | 0.801 | 1.567 |
| C2SDG | 2021 | 22.001 | 7.972 |
| ATTENTION SWIN U-NET | 2022 | 46.91 | 14.181 |
| MALUNet | 2022 | 0.175 | 0.083 |
| UNeXt-S | 2022 | 0.253 | 0.104 |
| EGE-UNet | 2023 | 0.053 | 0.072 |
| VM-UNet | 2024 | 27.427 | 4.112 |
| VM-UNet v2 | 2024 | 22.771 | 4.400 |
| LightM-UNet | 2024 | 0.403 | 0.391 |
| UltraLight VM-UNet | 2024 | 0.049 | 0.065 |
| UCM-Net (Baseline) | 2024 | 0.047 | 0.045 |
| MUCM-Net (1-patch) | 2024 | 0.139 | 0.064 |
| MUCM-Net (2-patch) | 2024 | 0.100 | 0.059 |
| MUCM-Net (4-patch) | 2024 | 0.081 | 0.057 |
| MUCM-Net (8-patch) | 2024 | 0.071 | 0.055 |
GFLOPS: Giga Floating-point Operations Per Second. ↓: the lower the value, the better the performance