search

facebookresearch / Mask2Former

Code release for "Masked-attention Mask Transformer for Universal Image Segmentation"
MIT License
2.59k stars 388 forks source link

Unable to reproduce Cityscapes results #202

Open yxchng opened 1 year ago

yxchng commented 1 year ago

Using maskformer2_R50_bs16_90k.yaml config, I got mIoU 76.2628 for semantic seg, AP 33.7673 for instance seg and PQ 59.9523 for panoptic seg. Does the results for Cityscapes have large variance? Do I need to run many times to get lucky results?

Screenshot from 2023-06-02 13-17-58

5had3z commented 9 months ago

I get approximately the same results when simply running their code with their configuration, you are not alone...