PAGE: Towards Practical Human-level Gaze Target Estimation

2026-07-06

AI Summary

Current gaze estimation models struggle to match human-level performance, limiting their practical deployment. This work proposes PaGE, a novel approach that explicitly models the complex interactions between scene context and head features and leverages large-scale unlabeled data through knowledge distillation from a ViT-H+ teacher model to train a lightweight student network. The resulting model achieves human-level or superior performance on seven out of nine benchmark metrics—the first method to do so—and reduces the human–machine performance gap by over 60% on the remaining two. By simultaneously attaining high accuracy and model efficiency, PaGE enables real-world applications in robotics and consumer devices.

Abstract

Gaze target estimation, the task of predicting where a person is looking in a scene, is crucial to understanding human attention and intent. It is a challenging task that combines high-level understanding of global scene semantics and precise spatial reasoning using human appearance (e.g. pose, eye orientation). As a result, human-level performance remains elusive for existing models, limiting their practical application. To this end, we propose PaGE (Practical Gaze Estimator), a gaze estimation model that explicitly models the complex interaction between scene and head features. Using a PaGE model with a large ViT-H+ backbone as the teacher, we further distill student models with lighter backbones on a much larger and more diverse unlabeled dataset. The architectural improvements and novel training recipe allow PaGE to achieve state-of-the-art performance on several gaze estimation tasks, outperforming humans in 7 out of 9 metrics while reducing the human-AI gap by at least 60% in the remaining 2. The distilled student models retain most of the teacher's performance while being lightweight enough for practical deployment on robots and consumer devices. The code and model checkpoints are available at our project page.

Problem

• gaze target estimation
• human-level performance
• scene semantics
• spatial reasoning
• practical deployment

Innovation

• gaze target estimation
• scene-head interaction modeling
• knowledge distillation
• vision transformer
• human-level performance