Neural Network Eye Tracking:Determining Nine-Grid Regions & Application

Abstract

As technology advances, the development of mobile platforms has nearly reached saturation. AR/MR (Augmented or Mixed Reality) eyewear is anticipated to be the new type of product leading technological advancements in the next 5-10 years. Some AR/MR glasses on the market with specialized functions have already become lightweight and convenient to use. Combining AR/MR glasses with eye-tracking technology implies that the latest technological advancements will significantly impact the daily life interactions and communications of users who are unable to use conventional interaction modes. Compared to the existing eye-tracking technology in AR/MR glasses, such as infrared eye-tracking, employing machine learning will make this technology more robust and adaptive to different environments, as well as providing better compatibility during calibration and setup. This thesis will demonstrate the use of cameras combined with neural networks for eye-tracking analysis and application scenarios, and hypothesize on applying this technology to future AR/MR devices that will be compatible with this technology.