Project description

The duel between the pitcher and the batter is a key moment in every baseball game. Therefore, in order to become a professional baseball player, the batter must learn how to correctly analyze the trajectory of the ball and take a position that is comfortable for hitting.

A special object recognition application is of great help in this analysis. The solution allows recording, upload to the server, sort, and store important fragments of the training process for later viewing and analysis. Let us tell about this in more detail.

How it works

The purpose of the application is to automatically record the moment the ball reaches home base and the batter tries to hit it. At the moment the ball reaching the frame, the system recognizes it and starts recording the fragment. In real-time, the app provides a preliminary estimate of the batter’s pose and ball trajectory (high, medium, low). The recorded video is compressed and transmitted to the server, where further analysis using machine learning algorithms is possible. Shooting is carried out by the user from a mobile device, usually mounted on a tripod, in outdoor conditions.

Main functionality

The system consists of two parts:

  • A mobile application for recognizing objects in the play area, as well as shooting and streaming video clips in automatic mode.
  • The server part for loading, sorting, and storing recorded fragments and conducting deeper processing: recognizing the position and posture of players, analyzing the trajectory of the ball, and other information (optional).

Players and ball trajectory detection

Pose estimation

Video processing

Object detection clarification

Technologies used