pith. sign in

arxiv: 2604.24331 · v2 · submitted 2026-04-27 · 💻 cs.CV

An Affordable, Wearable Stereo-Eye-Tracking Platform

Alexander Zimmer , Yasmeen Abdrabou , Enkelejda Kasneci This is my paper

Pith reviewed 2026-05-08 04:27 UTC · model grok-4.3

classification 💻 cs.CV
keywords eye trackingwearablestereo visioninfrared camerasopen source hardware3D printingcalibrationglint detection
0
0 comments X p. Extension

The pith

An affordable wearable stereo eye-tracking platform uses four infrared cameras and open 3D-printable designs to support multiple tracking methods in one modular setup.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a wearable eye-tracking device assembled from common electronic parts and 3D-printed frames. Current commercial and open trackers restrict how researchers can develop or compare new algorithms. This system places four infrared cameras around the eyes plus illumination, adds an optional forward scene camera, and supplies software for calibration plus synchronized recording. The same hardware can therefore run stereo reconstruction, glint detection, or binocular methods without physical changes. All files are released publicly so others can build and extend the platform for their own experiments.

Core claim

We present an affordable, wearable stereo eye-tracking platform built from off-the-shelf and 3D-printable components. The system combines four infrared eye cameras, infrared illumination, an optional scene camera, and software support for calibration and synchronized data acquisition. By design, the platform supports multiple eye-tracking paradigms, including stereo, glint-based, and binocular approaches, within a single hardware configuration. Rather than optimizing for end-user robustness, the platform prioritizes modularity and extensibility for research use. This paper focuses on the hardware architecture and calibration pipeline and demonstrates the feasibility of the approach using a 3

What carries the argument

The stereo eye-tracking platform: a head-worn assembly of four infrared eye cameras, synchronized illumination, optional scene camera, and calibration software that lets the same device run stereo, glint, or binocular tracking interchangeably.

If this is right

  • The same device can be used to run and directly compare stereo reconstruction against glint-based methods.
  • Open release of designs and software removes the hardware barrier for groups wanting to test new eye-tracking algorithms.
  • Synchronized eye and scene video enables studies of natural gaze behavior outside the lab.
  • Modular construction lets researchers add or swap cameras and illumination without redesigning the entire headgear.
  • Calibration routines supplied with the platform reduce setup time for different users and head sizes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Direct side-by-side algorithm tests on identical hardware could become routine rather than requiring multiple separate trackers.
  • Community modifications to the open files might produce lighter or more comfortable versions over time.
  • The platform could serve as a base for adding other sensors such as EEG or head-motion trackers in future extensions.
  • If the design proves stable, similar low-cost stereo camera rigs might appear for other wearable sensing tasks like hand tracking.

Load-bearing premise

The prototype built from standard components actually produces usable synchronized eye images and can be calibrated without special manufacturing or expert tuning.

What would settle it

A calibration run on the prototype that yields no usable pupil or glint positions, or shows that the four cameras cannot be time-synchronized to within a few milliseconds, would show the hardware does not deliver the claimed research platform.

Figures

Figures reproduced from arXiv: 2604.24331 by Alexander Zimmer, Enkelejda Kasneci, Yasmeen Abdrabou.

Figure 2
Figure 2. Figure 2: Sample recordings of a cameras pair and the scene view at source ↗
Figure 1
Figure 1. Figure 1: Overview of the eye tracker prototype. The firmware for the ESP32S3 is that used by EyeTrackVR [Eye￾TrackVR 2026] with minor adaptations. It supports wireless or wired (UVC) transmission, with the latter being inherently more robust at the cost of portability. The achievable frame rate is approximately 45 FPS, precise frame timing is transmitted with every frame. The resolution is 240x240 pixels, which is … view at source ↗
Figure 3
Figure 3. Figure 3: Visualization of some of the calibration steps. view at source ↗
read the original abstract

Research on video-based eye-tracking has long explored stereo and glint-based methods, yet existing wearable eye trackers - both commercial and open-source - offer limited flexibility for algorithm development and comparative evaluation. We present an affordable, wearable stereo eye-tracking platform built from off-the-shelf and 3D-printable components that explicitly targets this gap. The system combines four infrared eye cameras, infrared illumination, an optional scene camera, and software support for calibration and synchronized data acquisition. By design, the platform supports multiple eye-tracking paradigms, including stereo, glint-based, and binocular approaches, within a single hardware configuration. Rather than optimizing for end-user robustness, the platform prioritizes modularity and extensibility for research use. This paper focuses on the hardware architecture and calibration pipeline and demonstrates the feasibility of the approach using a prototype implementation. All hardware designs and documentation are made openly available.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

1 major / 0 minor

Summary. The manuscript presents the design of an affordable, wearable stereo eye-tracking platform built from off-the-shelf components and 3D-printable parts. The system integrates four infrared eye cameras with IR illumination, an optional scene camera, and software for calibration and synchronized multi-camera data acquisition. It is explicitly designed to support multiple paradigms (stereo, glint-based, and binocular) within one hardware configuration, prioritizing modularity and extensibility for research use over end-user robustness. The paper focuses on the hardware architecture and calibration pipeline, and asserts that a prototype implementation demonstrates the feasibility of the approach, with all designs released openly.

Significance. If the prototype delivers usable tracking performance, the work would address a genuine gap by supplying a flexible, low-cost open hardware platform that enables algorithm development and direct comparisons across eye-tracking methods. The emphasis on modularity and the public release of designs and documentation are clear strengths that could accelerate research in video-based eye tracking.

major comments (1)
  1. [Abstract and prototype demonstration] Abstract and prototype demonstration section: The central claim that the platform is a viable research tool rests on the assertion that the prototype 'demonstrates the feasibility of the approach,' yet no quantitative metrics are reported (e.g., angular gaze error, precision under head motion, sampling rate, or comparisons to existing open or commercial systems). Without these data the viability assertion remains unsupported.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their constructive feedback and for recognizing the value of a modular, open-source platform for eye-tracking research. We address the single major comment below and will revise the manuscript to strengthen the prototype demonstration.

read point-by-point responses

  1. Referee: [Abstract and prototype demonstration] Abstract and prototype demonstration section: The central claim that the platform is a viable research tool rests on the assertion that the prototype 'demonstrates the feasibility of the approach,' yet no quantitative metrics are reported (e.g., angular gaze error, precision under head motion, sampling rate, or comparisons to existing open or commercial systems). Without these data the viability assertion remains unsupported.

    Authors: We agree that the current manuscript provides insufficient quantitative support for the feasibility claim. The paper's emphasis is on the hardware architecture, modularity across tracking paradigms, and calibration pipeline, with the prototype serving primarily to confirm that the design can be realized and operated for synchronized data capture. To address this gap, the revised manuscript will expand the prototype section with the following: the achieved frame rate and synchronization performance of the four-camera system; calibration residual errors for the stereo and glint-based setups; and a qualitative assessment of tracking stability under moderate head motion. We will also add a concise comparison table against other open-source wearable eye trackers regarding cost, component count, and supported methods. These additions will be drawn from data already collected during prototype development and will be presented with appropriate caveats about the platform's research-oriented rather than end-user focus. revision: yes

Circularity Check

0 steps flagged

No circularity: descriptive hardware platform paper with no derivations or fitted predictions

full rationale

This is a hardware and software platform description paper. The abstract and provided text focus on presenting an affordable wearable stereo eye-tracking system using off-the-shelf and 3D-printable components, including four IR cameras, illumination, optional scene camera, and calibration software. No equations, mathematical derivations, parameter fitting, predictions, or first-principles results are present that could reduce to inputs by construction. The contribution is the open design and feasibility demonstration via prototype, with no self-citation chains or ansatzes invoked for any claimed outcomes. The derivation chain is empty by nature of the paper type, making it fully self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

This is an applied engineering paper describing hardware assembly and a calibration pipeline. No free parameters, mathematical axioms, or invented entities are introduced; the work relies on standard off-the-shelf components and established eye-tracking principles.

pith-pipeline@v0.9.0 · 5451 in / 1011 out tokens · 52801 ms · 2026-05-08T04:27:07.691480+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

18 extracted references · 16 canonical work pages

  1. [1]

    2018), 2480–2497

    Develop- ment and Validation of a High-Speed Stereoscopic Eyetracker.Behavior Research Methods50, 6 (Dec. 2018), 2480–2497. doi:10.3758/s13428-018-1026-7 Andrew T. Duchowski

    work page doi:10.3758/s13428-018-1026-7 2018

  2. [2]

    Behavior Research Methods, Instruments, & Computers34, 4 (Nov

    A Breadth-First Survey of Eye-Tracking Applications. Behavior Research Methods, Instruments, & Computers34, 4 (Nov. 2002), 455–470. doi:10.3758/BF03195475 ETRA ’26, June 01–04, 2026, Marrakesh, Morocco Zimmer et al. Andrew T. Duchowski. 2017.Eye Tracking Methodology: Theory and Practice(3rd ed.). Springer Publishing Company, Incorporated. Shaharam Eivazi,...

    work page doi:10.3758/bf03195475 2002

  3. [3]

    InProceedings of the 2018 ACM Symposium on Eye Tracking Research & Applications

    An Inconspicuous and Modular Head-Mounted Eye Tracker. InProceedings of the 2018 ACM Symposium on Eye Tracking Research & Applications. ACM, Warsaw Poland, 1–2. doi:10.1145/3204493.3208345 EyeTrackVR

    work page doi:10.1145/3204493.3208345 2018

  4. [4]

    doi:10.1109/TBME.2005.863952 Dan Witzner Hansen and Qiang Ji

    General theory of remote gaze estimation using the pupil center and corneal reflections.IEEE Transactions on Biomedical Engineering53, 6 (2006), 1124–1133. doi:10.1109/TBME.2005.863952 Dan Witzner Hansen and Qiang Ji

    work page doi:10.1109/tbme.2005.863952 2006

  5. [5]

    doi:10.1109/TPAMI.2009.30 Benedikt Hosp, Shahram Eivazi, Maximilian Maurer, Wolfgang Fuhl, David Geisler, and Enkelejda Kasneci

    In the Eye of the Beholder: A Survey of Models for Eyes and Gaze.IEEE Transactions on Pattern Analysis and Machine Intelligence32, 3 (2010), 478–500. doi:10.1109/TPAMI.2009.30 Benedikt Hosp, Shahram Eivazi, Maximilian Maurer, Wolfgang Fuhl, David Geisler, and Enkelejda Kasneci

    work page doi:10.1109/tpami.2009.30 2010

  6. [6]

    doi:10.3758/s13428-019-01305-2 Moritz Kassner, William Patera, and Andreas Bulling

    RemoteEye: An Open-Source High-Speed Remote Eye Tracker: Implementation Insights of a Pupil- and Glint-Detection Algorithm for High-Speed Remote Eye Tracking.Behavior Research Methods52, 3 (June 2020), 1387–1401. doi:10.3758/s13428-019-01305-2 Moritz Kassner, William Patera, and Andreas Bulling

    work page doi:10.3758/s13428-019-01305-2 2020

  7. [7]

    InPro- ceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiq- uitous Computing: Adjunct Publication(Seattle, Washington)(UbiComp ’14 Ad- junct)

    Pupil: an open source platform for pervasive eye tracking and mobile gaze-based interaction. InPro- ceedings of the 2014 ACM International Joint Conference on Pervasive and Ubiq- uitous Computing: Adjunct Publication(Seattle, Washington)(UbiComp ’14 Ad- junct). Association for Computing Machinery, New York, NY, USA, 1151–1160. doi:10.1145/2638728.2641695 ...

    work page doi:10.1145/2638728.2641695 2014

  8. [8]

    IEEE Transactions on Visualization and Computer Graphics27, 5 (May 2021), 2757–

    EllSeg: An Ellipse Segmentation Framework for Robust Gaze Tracking. IEEE Transactions on Visualization and Computer Graphics27, 5 (May 2021), 2757–

    2021

  9. [9]

    arXiv:2007.09600 [cs] doi:10.1109/TVCG.2021.3067765 Kyle Krafka, Aditya Khosla, Petr Kellnhofer, Harini Kannan, Suchendra Bhandarkar, Wojciech Matusik, and Antonio Torralba

    work page doi:10.1109/tvcg.2021.3067765 2007

  10. [10]

    In: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)

    Eye Tracking for Everyone. In 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE, Las Vegas, NV, USA, 2176–2184. doi:10.1109/CVPR.2016.239 Päivi Majaranta and Andreas Bulling

    work page doi:10.1109/cvpr.2016.239 2016

  11. [11]

    The Lean 4 Theorem Prover and Programming Language

    Eye Tracking and Eye-Based Human– Computer Interaction. InAdvances in Physiological Computing, Stephen H. Fair- clough and Kiel Gilleade (Eds.). Springer London, London, 39–65. doi:10.1007/978- 1-4471-6392-3_3 OptiKey

    work page doi:10.1007/978-

  12. [12]

    arXiv:1712.08900 [cs] doi:10.1016/j.cviu.2018

    PuRe: Robust Pupil Detection for Real-Time Pervasive Eye Tracking.Computer Vision and Image Understanding170 (May 2018), 40–50. arXiv:1712.08900 [cs] doi:10.1016/j.cviu.2018. 02.002 Thiago Santini, Diederick C. Niehorster, and Enkelejda Kasneci

    work page doi:10.1016/j.cviu.2018 2018

  13. [13]

    InProceedings of the 11th ACM Symposium on Eye Tracking Research & Applications

    Get a Grip: Slippage-Robust and Glint-Free Gaze Estimation for Real-Time Pervasive Head- Mounted Eye Tracking. InProceedings of the 11th ACM Symposium on Eye Tracking Research & Applications. ACM, Denver Colorado, 1–10. doi:10.1145/3314111.3319835 S.-W. Shih and J. Liu

    work page doi:10.1145/3314111.3319835

  14. [14]

    2004), 234–245

    A Novel Approach to 3-D Gaze Tracking Using Stereo Cameras.IEEE Transactions on Systems, Man and Cybernetics, Part B (Cybernetics) 34, 1 (Feb. 2004), 234–245. doi:10.1109/TSMCB.2003.811128 Lech Swirski, Andreas Bulling, and Neil Dodgson

    work page doi:10.1109/tsmcb.2003.811128 2004

  15. [15]

    InProceedings of the Symposium on Eye Tracking Research and Applications(Santa Barbara, California)(ETRA ’12)

    Robust real-time pupil tracking in highly off-axis images. InProceedings of the Symposium on Eye Tracking Research and Applications(Santa Barbara, California)(ETRA ’12). Association for Computing Machinery, New York, NY, USA, 173–176. doi:10.1145/2168556.2168585 Lech Swirski and Neil Dodgson

    work page doi:10.1145/2168556.2168585

  16. [16]

    A Fully-Automatic, Temporal Approach to Single Camera, Glint-Free 3D Eye Model Fitting. (2013). Kang Wang and Qiang Ji

    2013

  17. [17]

    Pattern Recognition79 (2018), 216–227

    3D gaze estimation without explicit personal calibration. Pattern Recognition79 (2018), 216–227. doi:10.1016/j.patcog.2018.01.031 Xucong Zhang, Yusuke Sugano, and Andreas Bulling

    work page doi:10.1016/j.patcog.2018.01.031 2018

  18. [18]

    InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems

    Evaluation of Appearance- Based Methods and Implications for Gaze-Based Applications. InProceedings of the 2019 CHI Conference on Human Factors in Computing Systems. ACM, Glasgow Scotland Uk, 1–13. doi:10.1145/3290605.3300646

    work page doi:10.1145/3290605.3300646 2019