Semantic Segmentation on RainCityscape
82.8mIoUWeatherSeg
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| WeatherSegLabeled Data Ratio=FS, Backbone=ResNet-1012026.04 | 82.8 | — | |
| UniMatchLabeled Data Ratio=FS, Backbone=ResNet-1012026.04 | 79.9 | — | |
| WeatherSegLabeled Data Ratio=1/2, Backbone=ResNet-1012026.04 | 79.4 | — | |
| OursAdversarial Attack Loss=MSE Loss2022.03 | 75.92 | — | |
| WeatherSegLabeled Data Ratio=1/4, Backbone=ResNet-1012026.04 | 74.2 | — | |
| OursAdversarial Attack Loss=LPIPS Loss2022.03 | 72.12 | — | |
| Ours w/o DDAdversarial Attack Loss=MSE Loss2022.03 | 70.11 | — | |
| WeatherSegLabeled Data Ratio=1/8, Backbone=ResNet-1012026.04 | 68.3 | — | |
| Ours w/o RBAdversarial Attack Loss=MSE Loss2022.03 | 65.12 | — | |
| Ours w/o DDAdversarial Attack Loss=LPIPS Loss2022.03 | 63.52 | — | |
| WeatherSegLabeled Data Ratio=1/16, Backbone=ResNet-1012026.04 | 62.4 | — | |
| UniMatchLabeled Data Ratio=1/16, Backbone=ResNet-1012026.04 | 59.6 | — | |
| Ours w/o RBAdversarial Attack Loss=LPIPS Loss2022.03 | 59.04 | — | |
| A3-TTASegmentation Model=DeepLabV3+, Backbone=ResNet-101, Resolution=512x2562026.02 | 53 | 56.73 | |
| TENTSegmentation Model=DeepLabV3+, Backbone=ResNet-101, Resolution=512x2562026.02 | 50.72 | 54.59 | |
| Ours w/o ALAdversarial Attack Loss=MSE Loss2022.03 | 47.45 | — | |
| SARSegmentation Model=DeepLabV3+, Backbone=ResNet-101, Resolution=512x2562026.02 | 41.7 | 46.22 | |
| CoTTASegmentation Model=DeepLabV3+, Backbone=ResNet-101, Resolution=512x2562026.02 | 41.31 | 45.83 | |
| MPRNetAdversarial Attack Loss=MSE Loss2022.03 | 39.45 | — | |
| Source OnlySegmentation Model=DeepLabV3+, Backbone=ResNet-101, Resolution=512x2562026.02 | 37.53 | 42.01 | |
| Ours w/o ALAdversarial Attack Loss=LPIPS Loss2022.03 | 31.07 | — | |
| Ours w/o ATAdversarial Attack Loss=MSE Loss2022.03 | 30.62 | — | |
| MPRNetAdversarial Attack Loss=LPIPS Loss2022.03 | 24.51 | — | |
| Ours w/o ATAdversarial Attack Loss=LPIPS Loss2022.03 | 21.94 | — |