Researchers have developed a new gaze estimation model called PaGE (Practical Gaze Estimator) that aims to achieve human-level performance in predicting where a person is looking. The model explicitly addresses the interaction between scene and head features, and uses a novel training approach involving distillation from a large teacher model to lighter student models. PaGE has demonstrated state-of-the-art results, outperforming humans on most metrics and significantly closing the human-AI gap. The distilled models are designed for practical deployment on devices like robots and consumer electronics. AI
IMPACT This model could significantly improve human-computer interaction and enable more sophisticated AI applications in robotics and consumer devices by accurately interpreting human attention.
RANK_REASON The cluster describes a new research paper detailing a novel model for gaze target estimation.
AI-generated summary · Google Gemini · from 3 sources. How we write summaries →
