Semantic Segmentation on Cityscapes (CS) (Adaptation mIoU)
40.08mIoU (Reported)PØDA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| PØDASource domain=GTA5, Target prompt (P)=driving, Zero-shot domain adaptation=true2024.10 | 40.08 | — | — | — | |
| source-onlySource domain=GTA5, Target prompt (P)=driving, Zero-shot domain adaptation=true2024.10 | 36.38 | — | — | — | |
| CLIPstylerSource domain=GTA5, Target prompt (P)=driving, Zero-shot domain adaptation=true2024.10 | 31.5 | — | — | — | |
| PØDASource Dataset=GTA5, Shots=0-shot, Backbone=ResNet-50, Pre-training=CLIP, Segmentation Head=DeepLabv3+2024.10 | — | 36.38 | 40.08 | 3.7 | |
| SM-PPMSource Dataset=GTA5, Shots=1-shot, Backbone=ResNet-101, Pre-training=ImageNet, Segmentation Head=DeepLabv22024.10 | — | 30.48 | 39.32 | 8.84 |