Egocentric Gaze Estimation via Neck-Mounted Camera

2026-02-12

AI Summary

This work proposes and investigates the novel task of gaze estimation from a neck-mounted camera perspective, addressing a significant gap in existing research that has predominantly focused on head-mounted setups. To facilitate this study, the authors introduce the first eye-tracking dataset specifically curated for the neck-level viewpoint and develop a geometry-aware multi-task learning model based on Transformers, termed GLC. The model incorporates an auxiliary out-of-bounds gaze classification task and a geometry-aware loss function. Experimental results demonstrate that the out-of-bounds classification substantially improves gaze estimation accuracy, while joint learning from both head- and neck-mounted views yields no significant performance gain. This study establishes a crucial benchmark and offers valuable insights for future research in wearable gaze tracking.

Technology Category

• Computer Vision: Motion & Tracking
• Intelligent Robots: Multimodal Perception & Sensor Fusion
• Machine Learning: Multi-instance/Multi-view Learning

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Security and Privacy: Data transparency and provenance

Abstract

This paper introduces neck-mounted view gaze estimation, a new task that estimates user gaze from the neck-mounted camera perspective. Prior work on egocentric gaze estimation, which predicts device wearer's gaze location within the camera's field of view, mainly focuses on head-mounted cameras while alternative viewpoints remain underexplored. To bridge this gap, we collect the first dataset for this task, consisting of approximately 4 hours of video collected from 8 participants during everyday activities. We evaluate a transformer-based gaze estimation model, GLC, on the new dataset and propose two extensions: an auxiliary gaze out-of-bound classification task and a multi-view co-learning approach that jointly trains head-view and neck-view models using a geometry-aware auxiliary loss. Experimental results show that incorporating gaze out-of-bound classification improves performance over standard fine-tuning, while the co-learning approach does not yield gains. We further analyze these results and discuss implications for neck-mounted gaze estimation.

Problem

• egocentric gaze estimation
• neck-mounted camera
• gaze estimation
• alternative viewpoints

Innovation

• neck-mounted gaze estimation
• egocentric vision
• gaze out-of-bound classification
• multi-view co-learning
• transformer-based gaze model