Skip to main navigation Skip to search Skip to main content

Towards Low-Power, High-Frequency Gaze Direction Tracking with an Event-Camera

  • Yvonne Vullers
  • , Luna Gava (Corresponding author)
  • , Arren Glover
  • , Chiara Bartolozzi

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

18 Downloads (Pure)

Abstract

Eye tracking is used in smart eye-wear to understand the gaze direction of the users’ eyes and must be accurate and with low-latency for pleasant user experience, and low-power for extended use time in a mobile device. However, visual eye tracking using traditional cameras is not ideal given the typical dynamics of eye-motion. The fixed sample rate (fps) is either under-sampling (lower accuracy) during fast eye motion during a saccade or oversampling (wasted computation) during gaze fixations when the eye-ball pose is stationary. An alternative technology, event-cameras, respond only to lighting change and therefore the visual signal adapts to the changing eye dynamics. High-frequency tracking can be achieved during saccades, while redundant processing is avoided during stationary gaze. In addition, due to small data packets, low-latency is also achieved. We present an event-camera-only eye-tracking proof-of-concept using a 5-DoF deformation model and demonstrate strong potential for accurate, low-latency (400 Hz) eye-tracking on a publicly available dataset; demonstrating the first gaze direction (pitch and yaw) estimation with an event-camera. We highlight the challenges still required to be solved to bring event camera technology to commercial viability for eye-tracking.

Original languageEnglish
Title of host publicationComputer Vision – ECCV 2024 Workshops
Subtitle of host publicationMilan, Italy, September 29–October 4, 2024, Proceedings, Part XXIII
EditorsAlessio Del Bue, Cristian Canton, Jordi Pont-Tuset, Tatiana Tommasi
PublisherSpringer
Pages214-227
Number of pages14
ISBN (Electronic)978-3-031-91989-3
ISBN (Print)978-3-031-91988-6
DOIs
Publication statusPublished - 12 May 2025
EventWorkshops that were held in conjunction with the 18th European Conference on Computer Vision, ECCV 2024 - Milan, Italy
Duration: 29 Sept 2024 → 4 Oct 2024

Publication series

NameLecture Notes in Computer Science (LNCS)
Volume15645
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceWorkshops that were held in conjunction with the 18th European Conference on Computer Vision, ECCV 2024
Country/TerritoryItaly
CityMilan
Period29/09/24 → 4/10/24

Keywords

  • event-camera
  • eye-tracking
  • low-latency

Fingerprint

Dive into the research topics of 'Towards Low-Power, High-Frequency Gaze Direction Tracking with an Event-Camera'. Together they form a unique fingerprint.

Cite this