RoboCup Vision: A Shared Dataset for Object Detection in Robot Soccer
摘要
RoboCup is an international initiative that promotes research in robotics and artificial intelligence by creating soccer robots capable of competing against humans according to FIFA rules in 2050. One of the main challenges for these robots is to make decisions automatically. This requires them to identify the status of their surrounding robots, such as their teammates, the opponent robots, the ball, and the goal. To improve research and development efforts in the RoboCup community and to help train and evaluate future deep-learning models, a specialized dataset has been developed for the RoboCup Middle Size League (MSL) soccer robot competition. This dataset is recorded and labeled with specific details, and it combines images of the RCT Toulon team and actual RoboCup games. It is available at https://github.com/iutgeiitoulon/MSL_Vision.git and is under an Open Data License, which is expected to significantly speed up the RoboCup community’s development process.