Georgia Tech researcher forecasts 3D gaze in AR

Fiona Ryan's 3D gaze forecasting model predicts user eye movement in first-person environments, enabling AR devices to pre-render scenes before users look. Her method, detailed in the paper 'Forecasting 3D Scanpaths in Egocentric Video,' uses egocentric video to anticipate scanpaths in 3D spaces. This allows AR systems to proactively render interactive content, reducing latency and improving user experience. Ryan's research, presented at IEEE CVPR in Denver, advances real-time AR responsiveness by predicting where users will look next. The approach shifts AR from reactive to proactive, making interactions smoother and more intuitive for users wearing smart glasses.