Datasets

SPL teams have released a number of datasets suitable for machine learning. Most of these datasets are oriented towards vision tasks (classification, object detection, semantic segmentation, and pose detection) but sound recognition is also present. In the problem domain of robot soccer vision there are relatively few object classes to be handled but there are significant challenges related to the movement of the camera platform and motion blur, changes in scene illumination, and issues related to camera image quality (e.g. noise, limited dynamic range, color shifts). A significant aspect of the challenge is that computer vision solutions must run in real-time on an embedded platform (E.g. an Intel Atom E3485 for the Nao V6).

Outside of RoboCup, these datasets may prove useful to developers of embedded or edge device machine learning-based solutions.

2024

2023

2022

2021

2020

2019

  • Bernd Poppinga and Tim Laue. JET-Net: Real-time object detection for mobile robots. In Stephan Chalup, Tim Niemueller, Jackrit Suthakorn, and Mary-Anne Williams, editors, RoboCup 2019: Robot World Cup XXIII, volume 11531 of Lecture Notes in Artificial Intelligence, pages 227–240. Springer, 2019. Dataset available at https://b-human.informatik.uni-bremen.de/public/JET-Net/
  • Heinrich Mellmann, Benjamin Schlotter, and Philipp Strobel. Toward Data Driven Development in RoboCup. In RoboCup 2019: Robot Soccer World Cup XXIII, 2019. Lecture Notes in Computer Science, vol 11531. Springer, Cham. Dataset (annotated robot logs) available at https://www2.informatik.hu-berlin.de/~naoth/videolabeling/
  • Di Giambattista Valerio, Mulham Fawakherji, Vincenzo Suriani, Domenico D. Bloisi, and Daniele Nardi. On field gesture-based robot-to-robot communication with nao soccer players. In RoboCup 2019: Robot World Cup XXIII, Springer International Publishing, pp. 367-375, 2019. Dataset (non-verbal) available at http://www.dis.uniroma1.it/~labrococo/?q=node/459
  • Marton Szemenyei and Vladimir Estivill-Castro. ROBO: robust, fully neural object detection for robot soccer. In Stephan K. Chalup, Tim Niemüller, Jackrit Suthakorn, and Mary-Anne Williams, editors, RoboCup 2019: Robot World Cup XXIII [Sydney, NSW, Australia, July 8, 2019], volume 11531 of Lecture Notes in Computer Science, pages 309–322. Springer, 2019. [ DOI | https://doi.org/10.1007/978-3-030-35699-6\_24 ]. Dataset available at https://github.com/szemenyeim/ROBO
  • Ball detection dataset available at https://www.kaggle.com/berlinunitednaoth/tk3balldetectionrobocup2019sydney
  • Nao lower camera ball detection dataset that was auto-labeled available at https://b-human.informatik.uni-bremen.de/public/datasets/balldetector_lc/

2018

2017