Factor-Informed Uncertainty Distillation for Gaze Estimation

2026-07-22

AI Summary

This work addresses the challenge of reliably estimating prediction uncertainty in gaze estimation under unconstrained environments, where existing models suffer significant performance degradation due to poor confidence calibration. The authors propose a teacher–student framework that leverages interpretable image-quality factors—such as illumination, sharpness, eye visibility, and symmetry—to guide a lightweight student network in efficiently estimating uncertainty within a single forward pass. The teacher model explicitly models error-inducing factors using gradient-boosted decision trees and transfers this knowledge to the student via curriculum learning and ranking-based supervision. Evaluated on ETH-XGaze, Gaze360, and MPIIFaceGaze benchmarks, the method substantially improves uncertainty calibration, error-ranking correlation, and selective prediction performance, demonstrating particularly strong robustness in unconstrained settings.

Abstract

Deep gaze estimation works well in controlled capture but degrades in unconstrained settings, where systems must reject unreliable predictions. Single-pass uncertainty (e.g., heteroscedastic regression) infers uncertainty from pixels without explicit input-validity cues, while sampling based methods are often too costly for real time use. We propose Factor-Informed Uncertainty Distillation (FIUD), a teacher-student framework that aligns uncertainty with interpretable image-quality failure modes. A gradient-boosting teacher predicts expected gaze error from factors such as illumination, sharpness, eye visibility and symmetry; a neural student distills these signals via curriculum learning and ranking supervision into a lightweight single-pass uncertainty head. Across ETH-XGaze, Gaze360, and MPIIFaceGaze (>300k samples), FIUD improves uncertainty, error rank correlation and selective prediction versus deterministic and sampling-based baselines, with the largest gains in unconstrained settings.

Problem

• gaze estimation
• uncertainty estimation
• unconstrained settings
• reliable prediction
• image-quality factors

Innovation

• Uncertainty Estimation
• Gaze Estimation
• Knowledge Distillation
• Curriculum Learning
• Image Quality Factors