Dual-Path Enhancements in Event-Based Eye Tracking: Augmented Robustness and Adaptive Temporal Modeling

Hoang M. Truong, Vinh-Thuan Ly, Huy G. Tran, Thuan-Phat Nguyen, Tram T. Doan · 2025

Read Dual-Path Enhancements in Event-Based Eye Tracking: Augmented Robustness and Adaptive Temporal Modeling by Hoang M. Truong, Vinh-Thuan Ly, Huy G. Tran, Thuan-Phat Nguyen, Tram T. Doan (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2504-09960.txt.

Description

Event cameras watch eyes by reporting change instead of full frames, a trick that suits augmented reality headsets. Hoang Truong and co-authors improve that tracking in a 2025 paper built on a lightweight network designed for edge devices. Real-world eye motion is abrupt and noisy, which defeats fragile models.

One upgrade is a tougher training pipeline that shifts events in time, flips them in space, and deletes some at random. The second is a hybrid model, KnightPupil, pairing an image backbone with a two-way recurrent unit and an adaptive state-space block. On a standard benchmark the system cut error notably on hard samples and placed well in a 2025 challenge. The work targets practical eye tracking for AR and VR gear.

Catalog record

CREATORHoang M. Truong, Vinh-Thuan Ly, Huy G. Tran, Thuan-Phat Nguyen, Tram T. Doan
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2504-09960

Browse the full library · About the Tower of Babel