Unsupervised Panoptic Segmentation on Cityscapes (val)
62.3PQSupervised
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| SupervisedTraining data=Cityscapes2025.04 | 62.3 | 81.8 | 75.1 | 62.4 | 62.1 | |
| CUPSTraining data=Cityscapes, Pseudo classes=272025.04 | 27.8 | 57.4 | 35.2 | 17.7 | 35.1 | |
| U2SegTraining data=COCO & ImageNet, Pseudo classes=800 + 272025.04 | 18.4 | 55.8 | 22.7 | 10.2 | 24.3 | |
| U2SegTraining data=COCO & ImageNet, Pseudo classes=800 + 27, note=results reported in [55]2025.04 | 17.6 | 52.7 | 21.7 | 8.4 | 24.2 | |
| DepthG + CutLERTraining data=Cityscapes & ImageNet, Pseudo classes=272025.04 | 16.1 | 45.4 | 21.1 | 3 | 25.7 |