Table 11 Evaluation results of different semantic segmentation models on BSData dataset.
Model | Backbone | mIoU\(\uparrow\) | Acc\(\uparrow\) | FPS\(\uparrow\) | Params\(\downarrow\) | Flops\(\downarrow\) |
|---|---|---|---|---|---|---|
U-Net50 | VGG-16 | 86.88 | 99.59 | 20.67 | 24.89 | 172.93 |
HRNet52 | HRNetV2-W18 | 89.04 | 99.67 | 23.51 | 9.64 | 14.29 |
DeepLabv3+53 | MobileNetV2 | 89.86 | 99.69 | 67.87 | 5.81 | 26.43 |
UPerNet54 | Swin-T | 90.18 | 99.74 | 15.42 | 58.94 | 179 |
SETR PUP38 | ViT-T | 89.13 | 99.68 | 5.32 | 308 | 309 |
Segmenter42 | ViT-S | 88.54 | 99.66 | 13.78 | 25.89 | 26.60 |
TransUnet40 | ViT-B | 88.35 | 99.61 | 25.60 | 93.23 | 32.24 |
SwinUnet55 | Swin-T | 88.89 | 99.60 | 24.69 | 41.34 | 34.79 |
Segformer46 | MiT-B0 | 89.01 | 99.67 | 75.37 | 3.72 | 5.19 |
Ours | MiT-B0 | 91.41 | 99.75 | 71.02 | 3.60 | 2.35 |
