Badminton Swing Form Correction through Computer Analysis

• 

  Pose Estimation Application on Badminton Swing

I want to design a machine-learning model that can correct badminton swings to improve technique and ultimately, prevent injuries. Badminton is a fast-paced racket sport that requires sufficient technique to achieve both good accuracy and power in badminton shots. As a badminton player myself, some people are unable to get coaching from personal trainers due to financial issues. So, this machine-learning model will help guide people as a temporary substitute for a trainer to help them better their badminton swing form. Solving numerous errors, collecting data and training the data with different RNN architectures were the hardest challenges I faced. While previous research has focused on detecting the different stages of the swing or described the effect of pose estimation on instructional learning, I plan to use pose estimation and CLIP embeddings to focus on the technique of the swing to understand how a swing works. I think an emphasis on the technique of the swing will fit this task better than previous research work because of its sole focus on one badminton swing.

I will be applying for the ML+Science Track.

Built With

• python
• youtube