SSHT
"SSHT" (Semi-supervised Source Hypothesis Transfer.)
Paper: Learning Invariant Representation with Consistency and Diversity for Semi-supervised Source Hypothesis Transfer
If you use conda, just run the following:
conda create -n ssht python=3.10
conda activate ssht
conda install pytorch==1.12.1 torchvision==0.13.1 torchaudio==0.12.1 cudatoolkit=11.3 -c pytorch
pip install scikit-learn tqdm opencv-python pillow scipy
Dataset
Office-31 can be found here.
Office-Home can be found here.
DomainNet can be found here.
Visda-2017 can be found here.
Here, we provide the label lists for the above datasets, for UDA, SSDA (1 shot and 3 shot).
Train and Test
Please refer run_sh.md.
Citation
If you use this code for your research, please consider citing:
@article{wang2021learning,
title={Learning Invariant Representation with Consistency and Diversity for Semi-supervised Source Hypothesis Transfer},
author={Wang, Xiaodong and Zhuo, Junbao and Cui, Shuhao and Wang, Shuhui},
journal={arXiv preprint arXiv:2107.03008},
year={2021}
}
Contact
If you have any question, contact to me: