shi-labs

/

oneformer_cityscapes_swin_large

like 2

Thanks for your great work.
I test your model on one image. And I find that semantic result is normal but instance/panoptic results are wrong. BTW, I test on Cityscapes image and all the results are correct.
Input Image:
Semantic result:
Instance result:
Panoptic result:

Cityscapes results as follows:
Semantic result:
Instance result:
Panoptic result: