EETnet: a CNN for Gaze Detection and Tracking for Smart-Eyewear

TL;DR

EETnet is a CNN designed for eye tracking with event-based cameras, capable of running on microcontrollers, enabling efficient and low-power gaze detection suitable for smart eyewear devices.

Contribution

The paper introduces EETnet, a novel CNN architecture optimized for event-based eye tracking on resource-constrained embedded devices.

Findings

EETnet can run on microcontrollers with limited resources.

The methodology includes training, evaluation, and quantization on a public dataset.

Two architecture versions: classification and regression.

Abstract

Event-based cameras are becoming a popular solution for efficient, low-power eye tracking. Due to the sparse and asynchronous nature of event data, they require less processing power and offer latencies in the microsecond range. However, many existing solutions are limited to validation on powerful GPUs, with no deployment on real embedded devices. In this paper, we present EETnet, a convolutional neural network designed for eye tracking using purely event-based data, capable of running on microcontrollers with limited resources. Additionally, we outline a methodology to train, evaluate, and quantize the network using a public dataset. Finally, we propose two versions of the architecture: a classification model that detects the pupil on a grid superimposed on the original image, and a regression model that operates at the pixel level.