Multi-Domain Referee Dataset: Enabling Recognition of Referee Signals on Robotic Platforms

Arka Mitra, Lukas Molnar, Jan-Nico Zaech, Yan Wu, Seonyeong Heo, Fisher Yu, Luc Van Gool

Keywords: Action Recognition, Robotic Soccer, Dataset, Domain Adaptation

TL;DR: A dataset for referee action recognition using robotic player, aimed towards making robocup more autonomous.

Abstract: Recognizing referee signals is crucial in human and RoboCup soccer games, where an emphasis currently lies on full robot autonomy through understanding referee signals. To advance towards this goal, we introduce the Multi-Domain Referee Dataset aimed at high-efficiency action recognition in RoboCup and examine the transfer between simulated and real domains in strongly structured settings. Our dataset includes 3,108 action sequences across four domains with over 183,000 images. Utilizing a recognition model on an Intel-Atom-based NAO robot, we demonstrate enhanced performance by merging real and synthetic data, and efficient learning of new signals with synthetic data updates, reducing acquisition efforts for future RoboCup rule modifications.

Submission Number: 6