Computer Science

Detection of eye contact with deep neural networks is as accurate as human experts

E. Chong, E. Clark-whitney, et al.

Discover a groundbreaking deep neural network model that automatically detects eye contact in egocentric video, achieving accuracy on par with human experts. This innovative research, conducted by Eunji Chong, Elysha Clark-Whitney, Audrey Southerland, Elizabeth Stubbs, Chanel Miller, Eliana L. Ajodan, Melanie R. Silverman, Catherine Lord, Agata Rozga, Rebecca M. Jones, and James M. Rehg, showcases precision and recall rates that could transform gaze behavior analysis in clinical and research contexts.

00:00

Playback language: English

Index

Abstract

This paper presents a deep neural network model for automatically detecting eye contact in egocentric video, achieving accuracy comparable to human experts. The model, trained on a large dataset of annotated images from diverse subjects (including those with Autism Spectrum Disorder), shows high precision (0.936) and recall (0.943) in validation, matching the performance of 10 human coders (mean precision 0.918, recall 0.946). This scalable, objective tool offers significant potential for gaze behavior analysis in clinical and research settings.

Publisher

Nature Communications

Published On

Oct 27, 2020

Authors

Eunji Chong, Elysha Clark-Whitney, Audrey Southerland, Elizabeth Stubbs, Chanel Miller, Eliana L. Ajodan, Melanie R. Silverman, Catherine Lord, Agata Rozga, Rebecca M. Jones, James M. Rehg

Related Publications

Explore these studies to deepen your understanding of the subject.

Engineering and Technology

Enhanced detection of threat materials by dark-field x-ray imaging combined with deep neural networks

T. Partridge, A. Astolfo, et al.

Psychology

Interoception is associated with the impact of eye contact on spontaneous facial mimicry

M. Imafuku, H. Fukushima, et al.

Food Science and Technology

A high sucrose detection threshold is associated with increased energy intake and improved post-prandial glucose response independent of the sweetness intensity of isocaloric sucrose solutions

V. Preinfalk, K. Schweiger, et al.

Computer Science

Diverse misinformation: impacts of human biases on detection of deepfakes on networks

J. Lovato, J. St-onge, et al.

Listen, Learn & Level Up

Over 10,000 hours of research content in 25+ fields, available in 12+ languages.

No more digging through PDFs, just hit play and absorb the world's latest research in your language, on your time.

listen to research audio papers with researchbunny