PupilNet v2.0: Convolutional Neural Networks for CPU based real time Robust Pupil Detection

TL;DR

This paper introduces PupilNet v2.0, a CNN-based system for real-time, robust pupil detection on CPU, significantly improving detection accuracy in challenging conditions with a novel two-stage CNN and a fast refinement method.

Contribution

It presents a new CNN architecture and a refinement technique for real-time pupil detection, outperforming existing algorithms on large, challenging datasets.

Findings

Detection rate improved by ~9% over state-of-the-art

Achieves ~7% detection accuracy increase on a 7ms CPU core

Evaluated on over 135,000 images, including new challenging data

Abstract

Real-time, accurate, and robust pupil detection is an essential prerequisite for pervasive video-based eye-tracking. However, automated pupil detection in realworld scenarios has proven to be an intricate challenge due to fast illumination changes, pupil occlusion, non-centered and off-axis eye recording, as well as physiological eye characteristics. In this paper, we approach this challenge through: I) a convolutional neural network (CNN) running in real time on a single core, II) a novel computational intensive two stage CNN for accuracy improvement, and III) a fast propability distribution based refinement method as a practical alternative to II. We evaluate the proposed approaches against the state-of-the-art pupil detection algorithms, improving the detection rate up to ~9% percent points on average over all data sets (~7% on one CPU core 7ms). This evaluation was performed on over 135,000 images: 94,000 images from the literature, and 41,000 new hand-labeled and challenging images contributed by this work (v1.0).