RiccardoRiccio/Fitness-AI-Trainer-With-Automatic-Exercise-Recognition-and-Counting

An extension of the previous 'Fitness-AI-Coach': a complete web application with real-time exercise recognition and counting. The exercise recognition model achieves 99% accuracy on the test set and 95% and 90% on two additional external datasets. || Give a star ⭐ to the repository if it was useful. Thank you! 😊

48
/ 100
Emerging

Combines MediaPipe pose estimation with BiLSTM neural networks to extract body landmarks and angles for exercise classification, processing sequential 30-frame features for temporal pattern recognition. The web interface built on Streamlit supports three operational modes: manual video upload analysis, real-time webcam tracking, and automatic exercise detection with rule-based rep counting using angle thresholds. Integrates OpenAI's GPT-3.5-turbo chatbot via LangChain for conversational fitness coaching within the same application.

132 stars.

No License No Package No Dependents
Maintenance 10 / 25
Adoption 10 / 25
Maturity 8 / 25
Community 20 / 25

How are scores calculated?

Stars

132

Forks

28

Language

Python

License

Last pushed

Feb 28, 2026

Commits (30d)

0

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/RiccardoRiccio/Fitness-AI-Trainer-With-Automatic-Exercise-Recognition-and-Counting"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.