[CVPR2025] SCAP: Transductive Test-Time Adaptation via Supportive Clique-based Attribute Prompting

The official repository for SCAP: Transductive Test-Time Adaptation via Supportive Clique-based Attribute Prompting.

Environment Preparation

conda create -n TTA python=3.11
conda activate TTA
pip install torch==2.4.0 torchvision==0.19.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu118
pip install -r requirement.txt

Data

You should download ImageNet-A, ImageNet-R, ImageNet-Sketch and ImageNet-V first.

Then your data directory should be organized in the following format:

• you_data_path
  • imagenet-a
  • imagenet-r
  • imagenet-sketch
  • imagenet-v2-matched-frequency-format-val

Evaluation

You can run the following commands for ImageNet-A/R/Sketch/V2 respectively:

bash scripts/A.sh

bash scripts/R.sh

bash scripts/K.sh

bash scripts/V.sh

Results

The results were obtained with a single NVIDIA 4090 GPU.

The Acc@1 comparison results against CLIP, few-shot learning methods and the latest test-time adaptation methods on 4 benchmark datasets.

Citation

If you find this code useful for your research, please cite our paper.

@article{zhang2025scap,
  title={SCAP: Transductive Test-Time Adaptation via Supportive Clique-based Attribute Prompting},
  author={Zhang, Chenyu and Xu, Kunlun and Liu, Zichen and Peng, Yuxin and Zhou, Jiahuan},
  journal={arXiv preprint arXiv:2503.12866},
  year={2025}
}

Acknowledgement

Our code is based on the PyTorch implementation of TPT and DART.

Contact

For any questions, feel free to contact us (xkl@stu.pku.edu.cn).

Welcome to our Laboratory Homepage (OV³ Lab) for more information about our papers, source codes, and datasets.