zhanghang1989 / PyTorch-Encoding

A CV toolkit for my papers.
https://hangzhang.org/PyTorch-Encoding/
MIT License
2.04k stars 450 forks source link

Support for CFNet (CVPR 2019) #269

Open netw0rkf10w opened 4 years ago

netw0rkf10w commented 4 years ago

Hi Hang,

Congratulations on your awesome work! I am really impressed!

I would like to know if you could share the code and pretrained models for your "Co-Occurrent Features in Semantic Segmentation" paper. (You've made all your implementations available, including the recent ResNeSt, but this CFNet seems to be an exception currently.)

I am looking for a strong semantic segmentation baseline in PyTorch, and according to this page your CFNet seems to be the best one.

Thank you very much in advance!

zhanghang1989 commented 4 years ago

Thanks for reminding me. I just recently switch back to PyTorch. I will provide pretrained models for CFNet soon.

Btw, I am also training a stronger model using ResNeSt. Please check back in about 1~2 weeks. Stay tuned.

netw0rkf10w commented 4 years ago

@zhanghang1989 That's great! I am looking forward to the updates. Thanks for the prompt reply!

soroushmbk commented 3 years ago

Hi

thanks for the great code and papers. I am interested to work on your "Co-Occurrent Features in Semantic Segmentation" paper. did you publish the code?

the atten.py file in models directory has similar modules with CFNet. What's the relation between this file and CFNet?

best regards

zhanghang1989 commented 3 years ago

Yes, atten.py is the code for CFNet. Using the hyper-parameters provided in the paper should reproduce the results. Note that the learning rate may be half of the regular settings for some datasets.

soroushmbk commented 3 years ago

thank you very much.