Gaze estimation databases: A comprehensive survey

Rawdha Karmi

doi:10.1016/j.engappai.2026.115829

Gaze estimation databases: A comprehensive survey

2026-08-10

0

PRE

AI

DOI：10.1016/j.engappai.2026.115829

Original

Save

Abstract

En 中文

Accurate gaze estimation is essential not only in psychology, cognitive neuroscience, and Human-Computer Interaction, but there continues to be difficulty in estimating the direction of gaze with accuracy due to head pose variation and real-world conditions. This survey of gaze estimation gives a complete account of current methods, datasets, and recent developments related to gaze estimation; more specifically, an emphasis on gaze estimation methods based on deep learning approaches is provided. The survey also illustrates that high-quality datasets are essential for improving algorithm performance and discusses ongoing challenges, e.g., environmental noise, variations in head pose; reduced resolution of images. The databases utilized for gaze estimation are presented in detail and classified based on public/private; structured/unstructured; static/dynamic format, with an examination of their pros and cons. A review of how recent deep learning techniques have resulted in increased accuracy and robustness of gaze estimation systems is provided. Finally, the survey aims to draw attention to the necessity for consistent evaluation metrics and reporting methods so that comparisons among studies can be made fairly and to help direct future research in this area.

Gaze estimation databases: A comprehensive survey

Abstract

Journal

Organization

Cited Papers