Machine Learning-based optimization of physical activity
| Title | Machine Learning-based optimization of physical activity |
|---|---|
| Summary | The project should be able to detect the difference between the current exercise and a reference version of the exercise. |
| Keywords | |
| TimeFrame | Spring 2025 |
| References | |
| Prerequisites | Machine Learning courses |
| Author | |
| Supervisor | Cristofer Englund, Kevin Hernandez Diaz |
| Level | Master |
| Status | Open |
Detect physical level of a person during physical activity based on images or video data. The project should focus on a few (3-7) excersices. Excersices could include weight lifting and yoga postures.
The project should focus on detecting a few key performance indicators eg. straightness of the back, alignment of arms and legs.
The idea is to use pose estimation models (yolopose or mediapipe) to extract position of the joints and track the movement. Then give feedback based on the detected quality of the exercise and physical level of the user.
The project should be able to detect the difference between the current exercise and a reference version of the exercise.
Yoga82 image data set could be used for the yoga posture direction. At Kaggle there are gym exercise datasets available for weight lifting images.
