Related Experiment Video
Updated: Nov 12, 2025
08:25
Combining Eye-tracking Data with an Analysis of Video Content from Free-viewing a Video of a Walk in an Urban Park Environment
Published on: May 7, 2019
9.3K
Human Eye Movements Reveal Video Frame Importance.
Zheng Ma1, Jiaxin Wu2, Sheng-Hua Zhong2
1The Smith-Kettlewell Eye Research Institute.
Computer
|March 22, 2021
Summary
Eye movement patterns reveal important video frames, improving video summarization. This study introduces a deep learning model using gaze data for more effective video content selection.
Area of Science:
- Computer Science
- Cognitive Science
- Human-Computer Interaction
Background:
- Human eye movements provide spatial information in images and videos.
- Video summarization aims to reduce video length while preserving the core narrative.
- Understanding viewer attention is crucial for effective video summarization.
Purpose of the Study:
- To investigate if eye movement patterns correlate with frame importance in videos.
- To develop a novel video summarization model incorporating gaze data.
- To evaluate the impact of eye movement information on summarization performance.
Main Methods:
- Recorded eye movements of subjects viewing videos from the SumMe dataset.
- Analyzed gaze consistency between selected and unselected video frames.
- Developed a multi-stream deep learning model integrating frame attributes and eye movement data.
Main Results:
- Higher gaze consistency was observed for frames selected in video summaries.
- The deep learning model incorporating gaze data outperformed models using only physical frame attributes.
- Eye movement data significantly enhanced the accuracy of video summarization.
Conclusions:
- Eye movement patterns reflect cognitive processing and selection of important video information.
- Gaze-based insights offer an innovative approach to improving automated video summarization algorithms.
- Integrating human attention data can lead to more relevant and effective video summaries.
