search

charlesCXK / TorchSemiSeg

[CVPR 2021] CPS: Semi-Supervised Semantic Segmentation with Cross Pseudo Supervision
MIT License
504 stars 74 forks source link

Weight decay difference between paper and config #29

Closed JihwanEom closed 3 years ago

JihwanEom commented 3 years ago

Hi, thank you for releasing awesome work!

I have a question about weight decay value config on resnet50 - pascal voc - 1/8 labels.

The paper said "The momentum is fixed as 0.9 and the weight decay is set to 0.0005(5e-4)"

But in config file, the weight decay value defined with 1e-4. """ In TorchSemiSeg/exp.voc/voc8.res50v3+2B.CPS+2BCutMix/config.py ... line 106: C.weight_decay = 1e-4 """

Which value should I follow? Thank you!

charlesCXK commented 3 years ago

Hi, we use weight_decay = 1e-4 in our experiments. The number in the paper is a typo. We also run experiments using weight_decay = 5e-4, the result is ~73.0% for voc8.res50v3+.CPS, so it will not influence the performance a lot.

JihwanEom commented 3 years ago

Noted. I'll close this issue.