FlatTrack: Eye-tracking with ultra-thin lensless cameras

TL;DR

This work introduces FlatTrack, a near-eye lensless gaze-tracking framework using a mask-based PhlatCam to achieve ultra-flat form factors suitable for AR/VR wearables. It combines a two-stage pipeline (lensless scene reconstruction followed by a gaze-regressor) and a large, real dataset of ~20k paired lensless measurements with ground-truth gaze directions. Through extensive experiments, the authors demonstrate that lensless gaze estimation can reach accuracy comparable to conventional lens-based trackers while delivering real-time performance, and they analyze the trade-offs between reconstruction methods and gaze estimators. The study also shows minimal loss when transitioning from lensed to lensless imaging in simulated cross-dataset evaluation, underscoring the practicality and potential privacy advantages of lensless eye-tracking for consumer devices.

Abstract

Existing eye trackers use cameras based on thick compound optical elements, necessitating the cameras to be placed at focusing distance from the eyes. This results in the overall bulk of wearable eye trackers, especially for augmented and virtual reality (AR/VR) headsets. We overcome this limitation by building a compact flat eye gaze tracker using mask-based lensless cameras. These cameras, in combination with co-designed lightweight deep neural network algorithm, can be placed in extreme close proximity to the eye, within the eyeglasses frame, resulting in ultra-flat and lightweight eye gaze tracker system. We collect a large dataset of near-eye lensless camera measurements along with their calibrated gaze directions for training the gaze tracking network. Through real and simulation experiments, we show that the proposed gaze tracking system performs on par with conventional lens-based trackers while maintaining a significantly flatter and more compact form-factor. Moreover, our gaze regressor boasts real-time (>125 fps) performance for gaze tracking.

Figures (6)

• Figure 1: FlatTrack Gaze Estimation. (a) PhlatCam lensless camera allows development of small form-factor gaze tracker. (b) We propose a large dataset of paired lensless captures and gaze directions, which we use to evaluate the efficacy of (c) various lensless gaze estimation techniques.
• Figure 2: PhlatCam Imaging Pipeline. (a) Shows the forward imaging process and the lensless capture. Note that the capture doesn't resemble the scene it is imaging. (b) Reconstructing the scene back involves solving an inverse problem computationally.
• Figure 3: FlatTrack Dataset.(a) The capture setup used to collect the FlatTrack dataset. Note the small ($\sim 4cm$) distance between the camera and the eye. (b)The 15x15 grid on the monitor used as stimulus points to capture data has an FoV of 53.03 degree and 29.80 degree along x & y axis. The pixel distance along x and y directions are 121.3 and 65.3 pixels respectively. (c) FlatNet reconstruction of 6 images each across the grid from 4 subjects are displayed
• Figure 4: Lensless Gaze Estimation Pipeline. We follow a two-stage approach. In the first stage, we use a fixed lensless reconstruction algorithm to obtain the scene estimate. In the second stage, a gaze regression neural network predicts the gaze vector given the reconstruction. Using the loss function, the gaze regressor is updated while the reconstruction algorithm is frozen.
• Figure 5: Different Lensless Reconstruction Methods. (a) FlatNet reconstruction. (b) Wiener deconvolution. FlatNet provides cleaner reconstruction.