Project information

• Category: Machine Learning
• Client: University of Kelaniya

The aim of this project is to explore the possibilities of using human pose detection to identify ideally batting form of players.

This is done using the Mediapipe library.The code extract the landmarks (poses) of the traget in the video and compares it with the ideal pose landmarks for that selected pose.

The ideal pose is extracted from a set of videos which contain the ideal landmark locations. Next the angles between selected landmark poses are calculated. An average of the landmark angles are used to compare with the target subjects landmark angles. Here cosine similarity is used to calculate the similarity between each landmark angle. Large deviation are then ploted out to identify the batting form performance.