A do-it-yourself computer vision based robotic ball throw trainer

Abstract

We demonstrate a self-training system for sports involving throwing a ball. We design a do-it-yourself (DIY) machinery that can be assembled using off-the-shelf items and integrates computer vision to visually track the ball throw accuracy. In this work, we demonstrate a system that can identify if the ball went through the hoop and approximately in which of the hoop's inner region. We envision that this preliminary design sets the foundation for a complete DIY sports IoT system that involves a hoola hoop, RaspberryPi, PiCamera and a LED strip, along with advanced ball placement and dynamics tracking.