Object detection for KRSBI robot soccer using PeleeNet on omnidirectional camera

Abstract

Kontes Robot Sepak Bola Indonesia (KRSBI) is an annual event for contestants to compete their design and robot engineering in the field of robot soccer. Each contestant tries to win the match by scoring a goal toward the opponent's goal. In order to score a goal, the robot needs to find the ball, locate the goal, then kick the ball toward goal. We employed an omnidirectional vision camera as a visual sensor for a robot to perceive the object's information. We calibrated streaming images from the camera to remove the mirror distortion. Furthermore, we deployed PeleeNet as our deep learning model for object detection. We fine-tuned PeleeNet on our dataset generated from our image collection. Our experiment result showed PeleeNet had the potential for deep learning mobile platform in KRSBI as the object detection architecture. It had a perfect combination of memory efficiency, speed and accuracy.