Analysis of Human Attentions for Face Recognition on Natural Videos and Comparison with CV Algorithm on Performance

Ragab Sayed, Mona; Yuting Lim, Rosary; Mandal, Bappaditya; Li, Liyuan; Hwee Lim, Joo; Sim, Terence

Abstract

Researchers have conducted many studies on human attentions and their eye gaze patterns for face recognition (FR), hoping to inspire new ideas to develop computer vision (CV) algorithms which perform like or even better than human. Yet, while these studies have been performed on still face images, human attention in natural videos have not been investigated. In this work, we conducted a human psychophysics experiment to study the human FR performance, attention and analyze eye gaze patterns on challenging natural videos. We compared human and machine FR performance in natural videos by applying a state-of-the-art deep convolutional neural network (dCNN). Our findings show a significant gap between machine and human on performance, especially for humans who achieve higher FR performance (highperformers). Learning from the cognitive capabilities of humans, in particular the high-performers, may aid in reducing this gap. Hence, we investigated persons’ attentions to unfamiliar faces in a challenging face recognition task, and studied their eye gaze patterns. We propose a new and effective computational approach for eye gaze pattern analysis on natural videos, which produces reliable results and reduces the time and manual efforts needed. Our findings reveal that humans in general attend more to the regions around the center of the face than the other facial regions. The high-performers in particular reveal attention to the lower part of the face in addition to the centre regions.

Citation

Ragab Sayed, M., Yuting Lim, R., Mandal, B., Li, L., Hwee Lim, J., & Sim, T. (2017, March). Analysis of Human Attentions for Face Recognition on Natural Videos and Comparison with CV Algorithm on Performance. Presented at 2017 AAAI Spring Symposium, Stanford University, USA

• Presentation Conference Type: Conference Paper (published)
• Conference Name: 2017 AAAI Spring Symposium
• Start Date: Mar 27, 2017
• End Date: Mar 29, 2017
• Online Publication Date: Mar 29, 2017
• Publication Date: Mar 27, 2017
• Deposit Date: Nov 17, 2023
• Publisher: Association for the Advancement of Artificial Intelligence
• Book Title: No. 7: Science of Intelligence: Computational Principles of Natural and Artificial Intelligence
• Public URL: https://keele-repository.worktribe.com/output/636589
• Publisher URL: https://aaai.org/papers/15291-15291-analysis-of-human-attentions-for-face-recognition-on-natural-videos-and-comparison-with-cv-algorithm-on-performance/
• Related Public URLs: https://aaai.org/proceeding/07-spring-2017/; ; https://aaai.org/conference/spring-symposia/sss17/