GazePrior: Zero-Shot AR/VR Eye Tracking via Learned 3D Gaze Reconstruction
Researchers have developed GazePrior, a novel approach for zero-shot eye tracking in AR/VR applications. This method utilizes a learned 3D prior to model human eye distributions, enabling the synthesis of realistic and diverse training data. By leveraging this synthesized data, eye tracking models achieve higher accuracy and robustness compared to existing zero-shot techniques, eliminating the need for costly real-world data collection for new devices. AI
IMPACT Enables more realistic and cost-effective AR/VR experiences by improving eye-tracking technology.