CrowdFix: An Eyetracking Dataset of Real Life Crowd Videos

TL;DR

This paper introduces CrowdFix, a new eye-tracking dataset of crowd videos with varying densities, to improve understanding of visual attention in crowded scenes and evaluate saliency models.

Contribution

It provides a novel eye-tracking dataset of crowd videos with annotations and assesses current saliency models on this data for future improvements.

Findings

State-of-the-art models perform variably across crowd densities.

The dataset reveals challenges in current saliency modeling for crowded scenes.

CrowdFix offers a new benchmark for saliency research in complex environments.

Abstract

Understanding human visual attention and saliency is an integral part of vision research. In this context, there is an ever-present need for fresh and diverse benchmark datasets, particularly for insight into special use cases like crowded scenes. We contribute to this end by: (1) reviewing the dynamics behind saliency and crowds. (2) using eye tracking to create a dynamic human eye fixation dataset over a new set of crowd videos gathered from the Internet. The videos are annotated into three distinct density levels. (3) Finally, we evaluate state-of-the-art saliency models on our dataset to identify possible improvements for the design and creation of a more robust saliency model.