Private source
Body Pose Detection with Image Processing
Bachelor's thesis: pose classification with CNN + BiLSTM
A deep learning system that classifies specific poses such as yoga and predicts pose accuracy as a percentage. It works on time series of the x/y coordinates of 33 body keypoints extracted with MediaPipe.
- Period
- 2024 – 2025
- Role
- Gazi University · EEE
- Status
- Thesis project
- Platforms
- Desktop
33body keypoints
3synchronized views
Overview
It performs real-time skeleton tracking and pose recognition. The PyQt5 interface has three synchronized views: the live camera feed, the reference pose and a skeleton simulation. Users watch pose accuracy live; the target is sports training and physical rehabilitation.
Screens
Screenshots coming soon