Elix - Health & Fitness AI Platform

A collection of AI-powered health and fitness tools built with computer vision, machine learning, and natural language processing.

Projects

1. Rep Counter (rep-counter/)

Real-time exercise repetition counter using computer vision and neural networks. Tracks squats and pushups with form quality assessment.

How it works:

• Uses MediaPipe for pose detection (body keypoints from video)
• A trained PyTorch neural network predicts exercise depth (0-1 continuous value)
• A wave pipeline algorithm detects complete reps and rates form quality (GREEN/YELLOW/RED)

Key files:

• wave_pipeline.py - Core wave pattern detection and segmentation algorithms
• regressor_final_squats.py - Main inference engine for squat analysis
• pushup/ - Pushup-specific inference and models

Tech stack: Python, PyTorch, MediaPipe, OpenCV, Matplotlib

---

2. Fitness Tracker (fitness-tracker/)

Multi-component fitness application with AI-powered tracking and analysis.

Chatbot (chatbot/)

Streamlit-based fitness tracker with natural language exercise logging. Uses OpenAI GPT to parse inputs like "I did 3 sets of 10 pushups" and stores workout data in SQLite with analytics dashboards.

# Set your API key
export OPENAI_API_KEY="your-key-here"
streamlit run fitness_tracker.py

Pose Estimation (pose-estimation/)

Framework for evaluating pose estimation models (MediaPipe, MoveNet Lightning & Thunder) on exercise form analysis. Includes rep counting, form fault detection, and performance benchmarking.

Exercise Classifier (exercise-classifier/)

ML-based exercise form classification using MediaPipe keypoints and template matching.

Main Devansh (main-devansh/)

Structured pose estimation comparison framework with modular architecture for testing different models on squat analysis.

Tech stack: Python, Streamlit, OpenAI, MediaPipe, MoveNet, PyTorch, Plotly, SQLite

---

Setup

Each subproject has its own requirements.txt. Install dependencies per project:

cd <project-folder>
pip install -r requirements.txt

Environment Variables

The fitness tracker chatbot requires:

OPENAI_API_KEY=your-openai-api-key

License

Private repository.