KaiyangZhou / Dassl.pytorch

A PyTorch toolbox for domain generalization, domain adaptation and semi-supervised learning.
MIT License
1.23k stars 171 forks source link

Implementation of the Cross Domain Adaptive Clustering(Domain Adaptation) #44

Closed shreejalt closed 2 years ago

shreejalt commented 2 years ago

Hi, Thanks for your excellent repository for different DA/SSL/DG algorithms. I found the CDAC paper very interesting, and I have implemented the paper in your repo and would like to merge it with your main Dassl repository.

Link to the paper: Link The original implementation by the authors can be found from this link: Link

Citing the original paper.

@InProceedings{li2021cross,
    author    = {Li, Jichang and Li, Guanbin and Shi, Yemin and Yu, Yizhou},
    title     = {Cross-Domain Adaptive Clustering for Semi-Supervised Domain Adaptation},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2021},
    pages     = {2505-2514}
}
KaiyangZhou commented 2 years ago

Thanks very much! Your implementation looks clean.

Is it a CVPR'19 or CVPR'21 paper? The update in the readme file puts CVPR'19 by the way.

Also, is this re-implementation tested? Do you have any numbers on the two datasets?

shreejalt commented 2 years ago

Hi, Sorry from my side, I will edit it to CVPR'21. Yes, I have tested the implementation. I have the numbers on the digit5 dataset

shreejalt commented 2 years ago

Hi @KaiyangZhou , Any updates regarding the PR?

KaiyangZhou commented 2 years ago

Sorry, was very busy.

The code looks nice and ready to be merged.

One last check: could you please run linter.sh and fix any warnings related to this PR. Thanks!

shreejalt commented 2 years ago

Hi @KaiyangZhou I have done the linter.sh and committed the code again. I have also added the custom lr scheduler that the paper uses in their experiments.

Thanks alot

shreejalt commented 2 years ago

Hi @KaiyangZhou There is a small bug in the scheduler function that I just checked

Can I correct it and the you can again merge?

i will creat a new PR with the solved bug and then you can merge it

thanks and sorry for inconvenience.