This repository contains the code for the various simulations and model training regimes, as well as the trained models and the code for evaluating the model on the various real image data sets used and developed for the article Byrne, S.A., Maquiling, V., Nyström, M., Kasneci, E., & Niehorster, D.C. (submitted). LEyes: A Lightweight Framework for Deep Learning-Based Eye Tracking using Synthetic Eye Images.
When using the code or model in this repository in your work, please cite Byrne et al. (submitted).
For more information or questions, e-mail: sean.byrne@imtlucca.it / dcnieho@gmail.com. The latest version of this repository is available from www.github.com/dcnieho/Byrneetal_LEyes
The code and model are licensed under the Creative Commons Attribution-NonCommercial-ShareAlike 4.0 (CC NC-BY-SA 4.0) license.
Contents:
model_training: this folder contains the code for generating the various simulations used in the paper, as well as the code for training our models on these simulations.trained_model: this folder contains the trained models used for the evaluations reported in Byrne et al. (submitted).eval: this folder contains the code for running the model evaluations on real eye images (note that the high resolution set is not publicly available).results_EDS2020: this folder contains images showing our model performance on all 2605 annotated images from the EDS2020 dataset.datasets: If you wish to run theevalfolder, acquire the needed dataset and place it in a subfolder here.
N.B.: complete details of changes made are available on github
- initial release
By downloading this data set, you expressly agree to the following conditions of release and acknowledge the following disclaimers issued by the authors:
Data are available by permission of the authors. Use of data in publications, either digital or hardcopy, must be cited as follows: Byrne, S.A., Maquiling, V., Nyström, M., Kasneci, E., & Niehorster, D.C. (submitted). LEyes: A Lightweight Framework for Deep Learning-Based Eye Tracking using Synthetic Eye Images..
The authors shall not be held liable for any improper or incorrect use or application of the data provided, and assume no responsibility for the use or application of the data or interpretations based on the data, or information derived from interpretation of the data. In no event shall the authors be liable for any direct, indirect or incidental damage, injury, loss, harm, illness or other damage or injury arising from the release, use or application of these data. This disclaimer of liability applies to any direct, indirect, incidental, exemplary, special or consequential damages or injury, even if advised of the possibility of such damage or injury, including but not limited to those caused by any failure of performance, error, omission, defect, delay in operation or transmission, computer virus, alteration, use, application, analysis or interpretation of data.
No warranty, expressed or implied, is made regarding the accuracy, adequacy, completeness, reliability or usefulness of any data provided. These data are provided "as is." All warranties of any kind, expressed or implied, including but not limited to fitness for a particular use, freedom from computer viruses, the quality, accuracy or completeness of data or information, and that the use of such data or information will not infringe any patent, intellectual property or proprietary rights of any party, are disclaimed. The user expressly acknowledges that the data may contain some nonconformities, omissions, defects, or errors. The authors do not warrant that the data will meet the user's needs or expectations, or that all nonconformities, omissions, defects, or errors can or will be corrected. The authors are not inviting reliance on these data, and the user should always verify actual data.