Semi-Supervised Gaze Estimation via Disentangled Subspace Contrastive Learning

Appearance-based gaze estimation always suffers from poor generalization due to limited annotated samples and insufficient dataset diversity. Leading approaches adopt weakly supervised learning to generate large-scale pseudo-labeled data from unconstrained real-world scenarios, aiming to mitigate the domain shifts. In this work, we devise a simple yet effective semi-supervised learning architecture that leverages unlabeled data to enhance domain generalization, thereby reducing reliance on labor-intensive manual annotations. Our key insight is to impose Jacobian regularization to disentangle feature representations into discriminative subspaces dedicated to specific gaze components, such as pitch and yaw angles. We further exploit the intrinsic ordinal ranking within each subspace for contrastive learning, enabling the model to learn robust gaze representations from a small set of labeled samples and an abundance of unlabeled ones. This ultimately yields our Disentangled Subspace Contrastive Learning (DSCL) framework. Extensive experiments on multiple benchmarks verify that the proposed DSCL is plug-and-play, achieving competitive performance using only 20\%, 10\%, and even 5\% of the annotated data under both in-domain and cross-domain evaluation settings. The public code is available at \href{https://github.com/da60266/DSCL}{https://github.com/da60266/DSCL}.

Appearance-based gaze estimation models typically struggle with poor generalization, as limited labeled training data and insufficient dataset diversity severely hinder practical deployment. Existing methods mostly rely on weak supervision to generate pseudo-labels to alleviate domain discrepancy, yet they still fail to fully exploit unlabeled real-world data. This work introduces a flexible semi-supervised learning framework to address these issues. We propose a disentangled subspace contrastive learning method, which separates visual features into independent subspaces corresponding to different gaze components via Jacobian regularization. Combined with ordinal ranking-based contrastive learning, our method effectively learns robust gaze representations from scarce labeled data and massive unlabeled samples. Extensive experiments demonstrate that the plug-and-play DSCL framework achieves excellent in-domain and cross-domain performance with only a tiny fraction of annotated data, greatly reducing the annotation cost of gaze estimation tasks.