Abstract

Pupil center and pupil contour are two of the most important features in the eye-image used for video-based eye-tracking. Well annotated databases are needed in order to allow benchmarking of the available- and new pupil detection and gaze estimation algorithms. Unfortunately, creation of such a data set is costly and requires a lot of efforts, including manual work of the annotators. In addition, reliability of manual annotations is hard to establish with a low number of annotators. In order to facilitate progress of the gaze tracking algorithm research, we created an online pupil annotation tool that engages many users to interact through gamification and allows utilization of the crowd power to create reliable annotations \cite{artstein2005bias}. We describe the tool and the mechanisms employed, and report results on the annotation of a publicly available data set. Finally, we demonstrate an example utilization of the new high-quality annotation on a comparison of two state-of-the-art pupil center algorithms.

Highlights

  • Reliable annotated data sets are the cornerstones of the development of new algorithms in many disciplines, especially those related to computer vision and machine learning

  • In the domains of speaker recognition or machine translation (Greenberg et al, 2014; Przybocki et al, 2009), annual challenges are organized by independent bodies that create annotated data sets, to stimulate the research and push forward the state-of-the-art methods

  • This paper proposes a novel solution to the problem of pupil annotation by the implementation and evaluation of an online system that allows the crowdsourcing of the annotation data

Read more

Summary

Introduction

Reliable annotated data sets are the cornerstones of the development of new algorithms in many disciplines, especially those related to computer vision and machine learning. Available annotated data sets facilitate comparison and benchmarking of new algorithms, methods and approaches. In the domains of speaker recognition or machine translation (Greenberg et al, 2014; Przybocki et al, 2009), annual challenges are organized by independent bodies that create annotated data sets, to stimulate the research and push forward the state-of-the-art methods. Pupil-based measurements have broad and numerous applications in behavioral sciences. (de Greef et al, 2009) uses and pupil estimations, along other data, to trigger responses in an adaptive system.

Objectives
Results
Discussion
Conclusion

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call

Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.