Inspiration
The theme of this hackathon i.e to build a fitness application for physical and mental wellbeing inspired us in building this application.
What it does
The application helps in finding the correctness of a yoga posture by comparing it with an expert's yoga pose. When the desired pose is achieved by the yoga practitioner it blows a whistle to let the practitioner know it's correct.
How we built it
Time spent: 48 Hours Hackathon We built on top of MediaPipe Tech stack used includes Opencv and other python Libraries
Challenges we ran into
Initially we tried adapting PoseNet. However due to time constraints, we stuck with it. Issues with getting skeleton to work with open-cv, and lack of examples available. In the end we had to resort back to mediapipe to get it to work properly
Accomplishments that we're proud of
Playing sound to confirm that skeletons are matching with more than 90% accuracy.
What we learned
We tried comparing two skeletons and learned list square method for that. Two of our team members were newbie to ML but working 48 hours continuously and enjoyed. Learned about affine transformation.
What's next for Virtual Yoga Instructor
- Working on UI .
- Additional pages to select exercises.
Log in or sign up for Devpost to join the conversation.