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.
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
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
Download the checkpoints from here and place under the directory ./weights
weights
├── visda
├── officehome
├── domainnet126
Run the following scripts:
$sh scripts/run_visda.sh
$sh scripts/run_officehome.sh
$sh scripts/run_domainnet126.sh
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}
}