An Innovative Table Tennis Scoring System Using Deep Residual Network

The rise of various sports in recent years has made more and more people pay attention to the referee's judgment. Not only the players themselves, but also the fans will pay attention to the verdict. Fans don't want players or teams they support to lose due to referee's bias. World-class competition must emphasize fairness and justice. However, because of the subjective consciousness or negligence of the referee, it may lead to wrong judgment. This study proposes an innovative table tennis scoring system using deep residual network based on classified data and the deep learning technology. The aim is to improve the judgment of subjective consciousness and negligence, so as to score the table tennis competition through intelligence and automation. This study is mainly divided into three major steps in practice. The first step is collecting image data, including table tennis table, table tennis balls, people, background, etc., as training images, verification images, and test videos. The second step is training a reliable deep learning model, by using deep learning to infer the scoring situation of the table tennis competition. The third step is to use the embedded system integrated camera as the image monitoring of the table tennis competition, and combine the trained deep learning model with the microcontroller unit to control the scoring display to display the immediate score result.