Learning Spatio-Temporal Feature Representations for Video-Based Gaze Estimation

This article describes research focused on improving gaze estimation using video data. The core of the work likely involves developing methods to extract and utilize both spatial and temporal information from video sequences to enhance the accuracy of gaze prediction. The use of 'spatio-temporal' suggests the researchers are considering the evolution of gaze over time, not just a single frame analysis. The source, ArXiv, indicates this is a pre-print, meaning it's likely a research paper submitted for peer review.