S³VAADA: Submodular Subset Selection for Virtual Adversarial Active Domain Adaptation

ICCV 2021

Harsh Rangwani, Arihant Jain*, Sumukh K Aithal*, R. Venkatesh Babu
Video Analytics Lab, Indian Institute of Science, Bengaluru

[Project Webpage] [Paper]

TLDR

Obtain performance close to supervised learning using small amount of labelled data (~10%) in target domain, for adapting a model from source to target domain.

Results on Office-Home

Setup the requirements

Run the following commands to setup your environment

git clone https://github.com/val-iisc/s3vaada.git
cd s3vaada/
pip install -r requirements.txt
cd models/
wget https://download.pytorch.org/models/resnet50-19c8e357.pth
mv resnet50-19c8e357.pth resnet50.pth
cd ../

Dataset

Office-31

Run the following commands to download and preprocess Office-31 dataset

cd datasets/
sh get_office.sh
python preprocess_office31.py
cd ../

Office-Home

Run the following commands to download and preprocess Office-31 dataset

cd datasets/
sh get_office-home.sh
python preprocess_office-home.py
cd ../

Training

python main.py --name w2a-s3vaada --source webcam --target amazon

Citation

If you find our work useful cite our paper using the following BibTeX entry.

@InProceedings{Rangwani_2021_ICCV,
    author    = {Rangwani, Harsh and Jain, Arihant and Aithal, Sumukh K and Babu, R. Venkatesh},
    title     = {S3VAADA: Submodular Subset Selection for Virtual Adversarial Active Domain Adaptation},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2021},
    pages     = {7516-7525}
}