Semantic Segmentation on ACDC (Rain)
75.3mIoUCoDA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CoDABackbone=SegFormer, Scene Specificity=Scene-Agnostic2024.03 | 75.3 | — | — | — | |
| HRDABackbone=SegFormer, Scene Specificity=Scene-Agnostic2024.03 | 73.6 | — | — | — | |
| MICBackbone=SegFormer, Scene Specificity=Scene-Agnostic2024.03 | 72.3 | — | — | — | |
| DINOv2 + CDTrain. data=ACDC Rain2026.02 | 68 | — | — | — | |
| DINOv2 + CDTrain. data=CityScapes2026.02 | 66.9 | — | — | — | |
| SePiCoBackbone=SegFormer, Scene Specificity=Scene-Agnostic2024.03 | 66.1 | — | — | — | |
| CLOUDSEncoder=ConvNext-L, zero-shot Domain Adaptation=true2023.12 | 64.4 | — | — | — | |
| DINOv2Train. data=CityScapes2026.02 | 63.1 | — | — | — | |
| STABackbone=SegFormer, Scene Specificity=Scene-Agnostic2024.03 | 61.3 | — | — | — | |
| DINOv2Train. data=ACDC Rain2026.02 | 60.7 | — | — | — | |
| DAFormerBackbone=SegFormer, Scene Specificity=Scene-Agnostic2024.03 | 59.9 | — | — | — | |
| CLOUDSEncoder=ResNet-101, zero-shot Domain Adaptation=true2023.12 | 53.4 | — | — | — | |
| CoCoOp+CAKIBase Model=SEEM-Large2026.05 | 50.4 | — | — | — | |
| CLOUDSEncoder=ResNet-50, zero-shot Domain Adaptation=true2023.12 | 49.7 | — | — | — | |
| DiffTPT + HisTPTModel Scale=Large2024.10 | 49.7 | — | — | — | |
| TPT + HisTPTModel Scale=Large2024.10 | 49.4 | — | — | — | |
| HisTPTModel Scale=Large2024.10 | 49.1 | — | — | — | |
| CoOp+CAKIBase Model=SEEM-Large2026.05 | 48.7 | — | — | — | |
| DiffTPTModel Scale=Large2024.10 | 48.2 | — | — | — | |
| CoOpBase Model=SEEM-Large2026.05 | 48.1 | — | — | — | |
| TPTModel Scale=Large2024.10 | 47.9 | — | — | — | |
| CoCoOpBase Model=SEEM-Large2026.05 | 47.7 | — | — | — | |
| SEEMModel Scale=Large2024.10 | 47.4 | — | — | — | |
| SEEM-Large2026.05 | 47.4 | — | — | — | |
| ULDASource Domain=Cityscapes, Prompt=driving under rain, Backbone=CLIP-ResNet-50, Base Model=DeepLabv3+2024.04 | 44.94 | — | — | — | |
| PIDAAdaptation setting=1-shot, Target image usage=single target image2024.10 | 42.66 | — | — | — | |
| AdaINAdaptation setting=1-shot, Target image usage=single target image2024.10 | 42.61 | — | — | — | |
| PØDASource Domain=Cityscapes, Prompt=driving under rain, Backbone=CLIP-ResNet-50, Base Model=DeepLabv3+2024.04 | 42.31 | — | — | — | |
| PØDAAdaptation setting=0-shot, Prompt=driving under rain2024.10 | 42.31 | — | — | — | |
| PØDASource domain=Cityscapes (CS), Target prompt (P)=driving under rain, Zero-shot domain adaptation=true2024.10 | 42.31 | — | — | — | |
| PODAEncoder=ResNet-50, zero-shot Domain Adaptation=true2023.12 | 42.3 | — | — | — | |
| CLIPStylerzero-shot Domain Adaptation=true2023.12 | 38.7 | — | — | — | |
| source-onlySource Domain=Cityscapes, Prompt=driving under rain, Backbone=CLIP-ResNet-50, Base Model=DeepLabv3+2024.04 | 38.2 | — | — | — | |
| source-only2024.10 | 38.2 | — | — | — | |
| source-onlySource domain=Cityscapes (CS), Target prompt (P)=driving under rain, Zero-shot domain adaptation=true2024.10 | 38.2 | — | — | — | |
| DiffTPT + HisTPTModel Scale=Tiny2024.10 | 37.7 | — | — | — | |
| TPT + HisTPTModel Scale=Tiny2024.10 | 37.2 | — | — | — | |
| CLIPstylerSource Domain=Cityscapes, Prompt=driving under rain, Backbone=CLIP-ResNet-50, Base Model=DeepLabv3+2024.04 | 37.17 | — | — | — | |
| CLIPstylerSource domain=Cityscapes (CS), Target prompt (P)=driving under rain, Zero-shot domain adaptation=true2024.10 | 37.17 | — | — | — | |
| HisTPTModel Scale=Tiny2024.10 | 36.7 | — | — | — | |
| CoCoOp+CAKIBase Model=SEEM-Tiny2026.05 | 36.3 | — | — | — | |
| CoOp+CAKIBase Model=SEEM-Tiny2026.05 | 35.6 | — | — | — | |
| DiffTPTModel Scale=Tiny2024.10 | 35.3 | — | — | — | |
| CoCoOpBase Model=SEEM-Tiny2026.05 | 35.2 | — | — | — | |
| TPTModel Scale=Tiny2024.10 | 34.9 | — | — | — | |
| CoOpBase Model=SEEM-Tiny2026.05 | 34.2 | — | — | — | |
| SEEMModel Scale=Tiny2024.10 | 33.1 | — | — | — | |
| SEEM-Tiny2026.05 | 33.1 | — | — | — | |
| PØDASource Dataset=CS, Shots=0-shot, Backbone=ResNet-50, Pre-training=CLIP, Segmentation Head=DeepLabv3+2024.10 | — | 38.2 | 42.31 | 4.11 | |
| SM-PPMSource Dataset=CS, Shots=1-shot, Backbone=ResNet-101, Pre-training=ImageNet, Segmentation Head=DeepLabv22024.10 | — | 29.78 | 32.23 | 2.45 |