This is a research project I worked on during my senior year at UNC Graphics and Virtual Reality Group, advised by Prof. Henry Fuchs. We leverage higher-level Purkinje images to improve eye tracking accuracy for both vergence and accommodation in this work. I designed and fabricated the user-customized eyetracker prototype suited for eyeglasses-style AR and VR displays. I also collected and analyzed Purkinje image data for training with convolutional neural network (CNN).
This work was accepted to ISMAR 2020.