Facing reality: an industrial view on large scale use of facial expression analysis

  • Authors:
  • Gábor Szirtes;Dániel Szolgay;Ákos Utasi;Dániel Takács;István Petrás;Gábor Fodor

  • Affiliations:
  • Realeyes, Budapest, Hungary;Realeyes, Budapest, Hungary;Realeyes, Budapest, Hungary;Realeyes, Budapest, Hungary;Realeyes, Budapest, Hungary;Realeyes, Budapest, Hungary

  • Venue:
  • Proceedings of the 2013 on Emotion recognition in the wild challenge and workshop
  • Year:
  • 2013

Quantified Score

Hi-index 0.00

Visualization

Abstract

In spite of the advances in the field of automatic face analysis (including detection, verification, recognition or expression classification), unconstrained, realistic conditions still present a great challenge for most computer based methods. We discuss some of the main issues from the perspective of industrial use. In particular we highlight the importance of high quality training data and argue that we cannot "cut corners" by omnipotent models. As an example for industrial application we take the task of primary emotion recognition and present the steps (and dead-ends) that lead to a solution that is "good enough" for commercial use.