Weakly Supervised Contrastive Adversarial Training for Learning Robust Features from Semi-supervised Data

Official implementation of Weakly Supervised Contrastive Adversarial Training for Learning Robust Features from Semi-supervised Data, accepted by CVPR 2025.

Requirements

Dependencies

python==3.10
coloredlogs==15.0.1
matplotlib==3.10.6
numpy==2.3.2
pandas==2.3.2
Pillow==11.3.0
PyYAML==6.0.2
torch==2.6.0+cu126
torchattacks==3.5.1
torchvision==0.21.0+cu126
tqdm==4.67.1

Verified working hardware environment

• Ubuntu 20.04.6 LTS
• CUDA 12.6
• NVIDIA RTX 3090 Ti

Example usage

python tunner.py

Citation

If you find this code useful for your research, please consider citing the following paper:

@InProceedings{Zhang_2025_CVPR,
    author    = {Zhang, Lilin and Wu, Chengpei and Yang, Ning},
    title     = {Weakly Supervised Contrastive Adversarial Training for Learning Robust Features from Semi-supervised Data},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2025},
    pages     = {25718-25727}
}