Point-of-Gaze Estimation on Embedded Platform

This work presents the design and development of a point-of-gaze device. A prototype was developed using off-the-shelf components, such as infrared LEDs, an 18 MPx camera, Raspberry Pi 4 and a Coral Edge TPU. The system was used to study and evaluate two different point-of-gaze estimation techniques: one based on a 3D geometric eye model and another based on neural networks. Accuracy and processing time were analyzed for each method. The prototype showed promising results and could be used to include accessibility features on graphical user interfaces.