UniGaze: Towards Universal Gaze Estimation via Large-scale Pre-Training

TL;DR

UniGaze introduces a large-scale self-supervised pre-training approach for gaze estimation, significantly enhancing cross-domain generalization and reducing dependence on labeled data by leveraging diverse in-the-wild facial datasets.

Contribution

It is the first to apply large-scale self-supervised pre-training to gaze estimation, demonstrating improved cross-domain performance and establishing effective pretraining strategies.

Findings

Self-supervised pretraining improves cross-domain gaze estimation.

Semantic-task designed pretraining approaches fail for gaze estimation.

UniGaze outperforms existing models in diverse dataset evaluations.

Abstract

Despite decades of research on data collection and model architectures, current gaze estimation models encounter significant challenges in generalizing across diverse data domains. Recent advances in self-supervised pre-training have shown remarkable performances in generalization across various vision tasks. However, their effectiveness in gaze estimation remains unexplored. We propose UniGaze, for the first time, leveraging large-scale in-the-wild facial datasets for gaze estimation through self-supervised pre-training. Through systematic investigation, we clarify critical factors that are essential for effective pretraining in gaze estimation. Our experiments reveal that self-supervised approaches designed for semantic tasks fail when applied to gaze estimation, while our carefully designed pre-training pipeline consistently improves cross-domain performance. Through comprehensive experiments of challenging cross-dataset evaluation and novel protocols including leave-one-dataset-out and joint-dataset settings, we demonstrate that UniGaze significantly improves generalization across multiple data domains while minimizing reliance on costly labeled data. source code and model are available at https://github.com/ut-vision/UniGaze.