Subject-Based Domain Adaptation for Facial Expression Recognition (FG2024)

Muhammad Osama Zeeshan¹, Muhammad Haseeb Aslam¹, Soufiane Belharbi¹, Alessandro Lameiras Koerich¹, Marco Pedersoli¹, Simon Bacon², Eric Granger¹

¹LIVIA, Dept. of Systems Engineering, ETS Montreal, Canada
²Dept. of Health, Kinesiology & Applied Physiology, Concordia University, Montreal, Canada

Domain Adaptation Setting

Subject-based MSDA

Installation

Dependencies

Using conda

conda create --name <env> --file <this file>

Using pip

pip install -r req.txt

Biovid Pain and Heat Dataset

Biovid datasets PartA can be downloaded from here: (https://www.nit.ovgu.de/BioVid.html#PubACII17)

Training of Source Subjects and Adaptation to Target Domain

CUDA_VISIBLE_DEVICES=0 python methods/msbda.py --sub_domains_datasets_path=$BIOVID_DATASET_PATH --sub_domains_label_path=$BIOVID_DATASET_LABEL_PATH --pain_db_root_path=$BIOVID_ROOT_FOLDER_PATH

Citation

@article{zeeshan2023subject,
  title={Subject-Based Domain Adaptation for Facial Expression Recognition},
  author={Zeeshan, Muhammad Osama and Aslam, Muhammad Haseeb and Belharbi, Soufiane and Koerich, Alessandro L and Pedersoli, Marco and Bacon, Simon and Granger, Eric},
  journal={arXiv preprint arXiv:2312.05632},
  year={2023}
}