Data pretaining to Chapter 2 of the PhD dissertation: "Resource-efficient Gaze Estimation"

This dataset accompanies the paper "Resource-efficient Gaze Estimation via Frequency-domain Multi-task Contrastive Learning" (Chapter 2 of the PhD dissertation). This research proposes a resource-efficient framework for gaze representation learning to improve data efficiency and computational efficiency. We introduce the frequency-domain gaze estimation, which exploits the feature extraction capability and the spectral compaction property of discrete cosine transform to substantially reduce the computational cost of gaze estimation systems for both calibration and inference. Moreover, to overcome the data labeling hurdle, we design a contrastive learning-based framework for unsupervised gaze representation learning. This dataset contains codes to reproduce the results.

本数据集配套于刊载于博士学位论文第2章的论文《基于频域多任务对比学习的资源高效注视估计（Resource-efficient Gaze Estimation via Frequency-domain Multi-task Contrastive Learning）》。本研究提出了一种面向注视表征学习（gaze representation learning）的资源高效框架，旨在提升数据效率与计算效率。我们提出了频域注视估计（gaze estimation）方法，该方法利用离散余弦变换（discrete cosine transform）的特征提取能力与频谱紧致特性，大幅降低了注视估计系统在标定与推理阶段的计算开销。此外，为解决数据标注难题，我们设计了一种基于对比学习（contrastive learning）的无监督注视表征学习框架。本数据集附带可复现研究结果的代码。