search

Haochen-Wang409 / U2PL

[CVPR'22 & IJCV'24] Semi-Supervised Semantic Segmentation Using Unreliable Pseudo-Labels & Using Unreliable Pseudo-Labels for Label-Efficient Semantic Segmentation
Apache License 2.0
436 stars 61 forks source link

Cityscapes Result #92

Closed Ziwei-Niu closed 2 years ago

Ziwei-Niu commented 2 years ago

Hi, I am really interested in your paper and it's a great work. But when I try to reproduce the result on cityscapes dataset using Multiple Gpus, I meet a bug:

99f058eae8fa07a73f98f5ade6022a2

And I looked up this error online and found that it was caused by an array crossing the bounds, I checked the code and found the code below in loss_helper.py may cause this issue: image

But I have no idea how to fix it , I will appreciate it if you give me some advice, thank you.

Haochen-Wang409 commented 2 years ago

Hi, thanks for your attention of our work!

This code seems to belong to OHEM, which is simply borrowed from AEL.

You could try turning off OHEM in config.yaml.