Awesome

Data Normalization - Gaze Estimation

This is a repository for the code used to generate input data for the paper Investigation of Architectures and Receptive Fields for Appearance-based Gaze Estimation. The main code repository can be found at https://github.com/yunhanwang1105/GazeTech

This repository contains code for normalizing ETH-XGaze, MPIIFaceGaze, and Gaze360 datasets with single face in resolutions of 224x224 and 448x448, and multi-region crops of face and two eyes for resolution 224x224. The separate code for different datasets can be found in individual folders.

We refer to the paper 'Revisiting Data Normalization for Appearance-Based Gaze Estimation' for data normalization in the gaze estimation task. An updated code for that paper can be found here. We also re-used the part of the code from ETH-XGze repository and FAZE repository.

Attention

The coordinate system of the gaze labels of the original MPIIFaceGaze dataset is different from our data normalization repository. The coordinate system in our data normalization repository is aligned with the one in ETH-XGaze dataset. The difference between the two coordinate systems is that the pitch and yaw directions are opposite from the corresponding ones.

Citation

If you use the results, pre-trained models, or method from our arXiv paper, please cite it as:

@misc{wang2023investigation,
      title={Investigation of Architectures and Receptive Fields for Appearance-based Gaze Estimation}, 
      author={Yunhan Wang and Xiangwei Shi and Shalini De Mello and Hyung Jin Chang and Xucong Zhang},
      year={2023},
      eprint={2308.09593},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}

If you only use the data normalization part, please cite the data normalization paper:

@inproceedings{zhang2018revisiting,
  title={Revisiting data normalization for appearance-based gaze estimation},
  author={Zhang, Xucong and Sugano, Yusuke and Bulling, Andreas},
  booktitle={Proceedings of the 2018 ACM symposium on eye tracking research \& applications},
  pages={1--9},
  year={2018}
}