Home

Awesome

[ECCV 2024] CoDA: Instructive Chain-of-Domain Adaptation with Severity-Aware Visual Prompt Tuning

by Ziyang Gong, Fuhao Li, Yupeng Deng, Deblina Bhattacharjee, Xianzheng Ma, Xiangwei Zhu, Zhenming Ji

<div align="center"> <table> <tr> <td align="center"> <img src="images/sysu_logo.png" width="150px" height="150px"><br> <span style="font-size: 18px;"><b>Sun Yat-sen University</b></span> </td> <td align="center"> <img src="images/CPNT_logo.png" width="150px" height="150px"><br> <span style="font-size: 18px;"><b>CPNT Lab</b></span> </td> <td align="center"> <img src="images/EPFL_logo.png" width="150px" height="150px"><br> <span style="font-size: 18px;"><b>EPFL</b></span> </td> </tr> </table> </div>

🌟🌟🌟 Home

Here is the official project of 🎻CoDA. We are releasing the training code and dataset generated by ourselves in our paper.

CoDA is a UDA methodology that boosts models to understand all adverse scenes (☁️,☔,❄️,🌙) by highlighting the discrepancies between and within these scenes. CoDA achieves state-of-the-art performances on widely used benchmarks.

🔥🔥🔥 News

[2024-7-10]We have released our generated data samples. You can download from here.

<ins>[Baidu Netdisk]</ins>    <ins>[Google Drive]</ins>

[2024-7-2] We are delighted to inform that CoDA has been accepted by ECCV 2024 main conference 🎉🎉🎉!!!

[2024-3-8] We create the official project of CoDA and release the inference code.

Overview

night

PWC

PWC

PWC

PWC

PWC <a href="" target='_blank'><img src="https://visitor-badge.laobi.icu/badge?page_id=Cuzyoung.CoDA&left_color=%23DFA3CB&right_color=%23CEE75F"> </a>

<!-- ![visitors](https://visitor-badge.glitch.me/badge?page_id=Cuzyoung.CoDA&left_color=%23DFA3CB&right_color=%23CEE75F) -->

CoDA

ExperimentsmIoUCheckpoint
Cityscapes $\rightarrow$ ACDC72.6-
Cityscapes $\rightarrow$ Foggy Zurich60.9-
Cityscapes $\rightarrow$ Foggy Driving61.0-
Cityscapes $\rightarrow$ Dark Zurich61.2-
Cityscapes $\rightarrow$ Nighttime Driving59.2-
Cityscapes $\rightarrow$ BDD100K-Night41.6-

If you find this project useful in your research, please consider citing:

@article{gong2024coda,
  title={CoDA: Instructive Chain-of-Domain Adaptation with Severity-Aware Visual Prompt Tuning},
  author={Gong, Ziyang and Li, Fuhao and Deng, Yupeng and Bhattacharjee, Deblina and Zhu, Xiangwei and Ji, Zhenming},
  journal={arXiv preprint arXiv:2403.17369},
  year={2024}
}

Download Checkpoint

cd CoDA
python ./tools/download_ck.py

or you can manually download checkpoints from Google Drive.

Environment

conda create -n coda python=3.8.5 pip=22.3.1
conda activate coda
pip install -r requirements.txt -f https://download.pytorch.org/whl/torch_stable.html
pip install mmcv-full==1.3.7 -f https://download.openmmlab.com/mmcv/dist/cu110/torch1.7/index.html

Before run demo, first configure the PYTHONPATH, or you may encounter error like 'can not found tools...'.

cd CoDA
export PYTHONPATH=.:$PYTHONPATH

or directly modify the .bashrc file

vi ~/.bashrc
export PYTHONPATH=your path/CoDA:$PYTHONPATH
source ~/.bashrc

demo

python ./tools/image_demo.py --img ./images/night_demo.png --config ./configs/coda/csHR2acdcHR_coda.py --checkpoint ./pretrained/CoDA_cs2acdc.pth

Inference Steps

python ./tools/image_demo.py --img_dir ./acdc_dir --config ./configs/coda/csHR2acdcHR_coda.py --checkpoint ./pretrained/CoDA_cs2acdc.pth --out_dir ./workdir/cs2acdc

Traning Steps

python ./tools/train.py --config ./configs/coda/csHR2acdcHR_coda.py --work-dir ./workdir/cs2acdc