About the project

1. Introduction

• The importance of data analytics is growing in various elite sports field.
  • Many coaches have been giving feedback to individual players through data collected during matches, and designing best training programs accordingly.
  • Also, they devise the best strategy that can maximize the probability of winning by analyzing the data of opponents before each match.
• This kind of data analysis is very helpful for improving skills not only for elite players but also for amateur players.
• However, ordinary recreational players can not enjoy the benefits of data in the field of sports, because it is impossible for them to get these kinds of numerical data from their tennis match videos.
• I think computer vision can help improve amateur players’ overall sports experience quality by providing these kinds of numerical data at low cost.
• Since I am particularly interested in tennis among many sports, this project focused on how to obtain meaningful tennis data at a low cost.

2. Purpose & Subprojects

• The end result of this project is to extract basic statistics related to each player from tennis match clips: for example, the position of the players, the distance they run, the placement of their shots, etc.
• To do this, it is necessary to detect the players and the tennis ball in the match clip. -> tennis ball & players detection
• Also, since the coordinates of the detected balls and players must be converted to coordinates on the tennis court, tennis court detection is also necessary. -> tennis court detection
• After detecting the player, ball, and court, we need to convert the coordinates in the image to coordinates on the tennis court. -> Bird’s Eye View Transformation
• Also, since the distribution of players’ shots needs to be analyzed, it is necessary to learn a classification model that classifies players’ shots. -> Tennis shot classification
• That is, for this project, the following 5 sub-projects must be completed.
  • Tennis ball detection
  • Tennis players detection
  • Tennis court detection
  • Bird’s Eye view Transformation
  • Tennis shot classification