/pSTarC

Official code for the paper "pSTarC: Pseudo Source Guided Target Clustering for Fully Test-Time Adaptation"

Primary LanguagePythonMIT LicenseMIT

pSTarC

Official code for the paper

pSTarC: Pseudo Source Guided Target Clustering for Fully Test-Time Adaptation (WACV 2024) [Project-page] [arXiV]

Manogna Sreenivas, Goirik Chakrabarty and Soma Biswas.

Setup

Installation

All experiments were done using PyTorch 1.13 on NVIDIA A-5000 GPU. The environment can be setup as follows:

conda create -n pstarc
conda activate pstarc
conda install pytorch==1.13.0 torchvision==0.14.0 pytorch-cuda=11.7 -c pytorch -c nvidia
pip install scikit-learn matplotlib

Datasets

Download the VisDA, Office-Home, DomainNet-126 datasets and place it under data folder as follows:

data
├── visda
│   ├── train
│   ├── validation
├── officehome
│   ├── Art
│   ├── Clipart
│   ├── Product
│   ├── Realworld
├── domainnet126
│   ├── real
│   ├── sketch
│   ├── clipart
│   ├── painting

Checkpoints

Download the checkpoints from here and place under the directory ./weights

weights
├── visda
├── officehome
├── domainnet126

Test Time Adaptation

Run the following scripts:

VisDA

$sh scripts/run_visda.sh

Office-Home

$sh scripts/run_officehome.sh

DomainNet-126

$sh scripts/run_domainnet126.sh

Citation

If you use this code your work, please cite our paper

@inproceedings{sreenivas2024pstarc,
  title={pSTarC: Pseudo Source Guided Target Clustering for Fully Test-Time Adaptation},
  author={Sreenivas, Manogna and Chakrabarty, Goirik and Biswas, Soma},
  booktitle={Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision},
  pages={2702--2710},
  year={2024}
}