EyeTheia: A Lightweight and Accessible Eye-Tracking Toolbox. Pather, S., Martignène, N., Bugnet, A., Boutaleb, F., D’Hondt, F., & Maia, D. S. Lecture Notes in Computer Science, 16821 LNCS:526-541, Springer Science and Business Media Deutschland GmbH, 2027.
EyeTheia: A Lightweight and Accessible Eye-Tracking Toolbox [link]Paper  doi  abstract   bibtex   
We introduce EyeTheia, a lightweight and open deep learning pipeline for webcam-based gaze estimation, designed for browser-based experimental platforms and real-world cognitive and clinical research. EyeTheia enables real-time gaze tracking using only a standard laptop webcam, combining MediaPipe-based landmark extraction with a convolutional neural network inspired by iTracker and optional user-specific fine-tuning. We investigate two complementary strategies: adapting a model pretrained on mobile data and training the same architecture from scratch on a desktop-oriented dataset. Validation results on MPIIFaceGaze show comparable performance between both approaches prior to calibration, while lightweight user-specific fine-tuning consistently reduces gaze prediction error. We further evaluate EyeTheia in a realistic Dot-Probe task and compare it to the commercial webcam-based tracker SeeSo SDK. Results indicate strong agreement in left–right gaze allocation during stimulus presentation, despite higher temporal variability. Overall, EyeTheia provides a transparent and extensible solution for low-cost gaze tracking, suitable for scalable and reproducible experimental and clinical studies. The code, trained models, and experimental materials are publicly available (EyeTheia repository and Calypso experimental platform).
@article{Pather2027,
   abstract = {We introduce EyeTheia, a lightweight and open deep learning pipeline for webcam-based gaze estimation, designed for browser-based experimental platforms and real-world cognitive and clinical research. EyeTheia enables real-time gaze tracking using only a standard laptop webcam, combining MediaPipe-based landmark extraction with a convolutional neural network inspired by iTracker and optional user-specific fine-tuning. We investigate two complementary strategies: adapting a model pretrained on mobile data and training the same architecture from scratch on a desktop-oriented dataset. Validation results on MPIIFaceGaze show comparable performance between both approaches prior to calibration, while lightweight user-specific fine-tuning consistently reduces gaze prediction error. We further evaluate EyeTheia in a realistic Dot-Probe task and compare it to the commercial webcam-based tracker SeeSo SDK. Results indicate strong agreement in left–right gaze allocation during stimulus presentation, despite higher temporal variability. Overall, EyeTheia provides a transparent and extensible solution for low-cost gaze tracking, suitable for scalable and reproducible experimental and clinical studies. The code, trained models, and experimental materials are publicly available (EyeTheia repository and Calypso experimental platform).},
   author = {Stevenson Pather and Niels Martignène and Arnaud Bugnet and Fouad Boutaleb and Fabien D’Hondt and Deise Santana Maia},
   doi = {10.1007/978-3-032-31438-3_35/TABLES/3},
   isbn = {9783032314376},
   issn = {16113349},
   journal = {Lecture Notes in Computer Science},
   keywords = {Cognitive neuroscience,Deep learning,Eye tracking,Gaze estimation,Web-based experimentation},
   pages = {526-541},
   publisher = {Springer Science and Business Media Deutschland GmbH},
   title = {EyeTheia: A Lightweight and Accessible Eye-Tracking Toolbox},
   volume = {16821 LNCS},
   url = {https://link.springer.com/chapter/10.1007/978-3-032-31438-3_35},
   year = {2027}
}

Downloads: 0