DPA: Dual Prototypes Alignment for Unsupervised Adaptation of Vision-Language Models

This is the PyTorch code of the DPA paper.

Dataset Setup

Dataset paths are stored in dataset_catalog.json, which need to be modified to local paths. Please refer to the scripts from VISSL to download and prepare the datasets. The dataset labels are stored in classes.json.

Requirements

• PyTorch 1.10.0 or later
• timm 0.4.12
• tensorboardX
• ftfy

Training

Run the following command:

python train.py --dataset [name_of_dataset]

Citation

@InProceedings{Ali_2025_WACV,
    author    = {Ali, Eman and Silva, Sathira and Khan, Muhammad Haris},
    title     = {DPA: Dual Prototypes Alignment for Unsupervised Adaptation of Vision-Language Models},
    booktitle = {Proceedings of the Winter Conference on Applications of Computer Vision (WACV)},
    month     = {February},
    year      = {2025},
    pages     = {6083-6093}
}

Acknowledgements

Our code is based on MUST. We thank the authors for releasing their code.