# FitMentor AI - Form Checker (Alpha v0.1)

An AI-powered fitness coach that checks your workout form in real-time using computer vision.

Transferable Applications

While demonstrated with fitness/wellness use cases, this system's core technologies transfer to:

Manufacturing Quality Control:

• Assembly line worker ergonomics monitoring
• Repetitive motion analysis for injury prevention
• Real-time safety compliance verification

Healthcare & Physical Therapy:

• Patient rehabilitation progress tracking
• Gait analysis and fall risk assessment
• Post-surgery recovery monitoring

Sports Performance:

• Athletic technique analysis
• Injury prevention through biomechanical assessment
• Performance optimization coaching

## 🎯 Current Features

### Working Exercises:

- ✅ **Squat** - Depth & knee tracking analysis

- ✅ **Push-up** - Rep counter with body alignment check

- ✅ **Plank** - Hold timer with form monitoring

- ✅ **Lunge** - Front/back leg angle analysis

- ✅ **Deadlift** - Back safety monitor (critical for injury prevention)

### Technology:

- Real-time pose detection (MediaPipe)

- Angle calculation & biomechanics analysis

- Form scoring (0-100)

- Color-coded feedback (green/yellow/red)

## 🚧 Alpha Status - Known Issues

This is an early prototype. Tracking needs significant improvement:

**General Issues:**

- Angle detection needs smoothing/filtering

- Rep counting can be inconsistent

- Works best from specific camera angles

- No calibration for different body types

**Specific Exercise Issues:**

- **Squat:** Best from direct side view, sometimes false counts

- **Push-up:** Position detection oversensitive

- **Plank:** Timer can be inconsistent

- **Lunge:** Struggles detecting front vs back leg

- **Deadlift:** Back angle needs refinement

## 🎯 Improvement Roadmap

Week 2 Focus:

- [ ] Add angle smoothing/filtering

- [ ] Improve rep counting logic with state machines

- [ ] Add confidence thresholds

- [ ] Better "ready position" detection

- [ ] Calibration for body dimensions

Future:

- [ ] Support multiple camera angles

- [ ] Add more exercises

- [ ] Mobile app

- [ ] Personal AI coach (LLM integration)

## 🛠️ Setup

\# Clone repo

git clone https://github.com/Vamsikrishnv/fitmentor-form-checker.git

cd fitmentor-form-checker



\# Create virtual environment

python -m venv venv

source venv/bin/activate  # Windows: venv\\Scripts\\activate



\# Install dependencies

pip install -r requirements.txt

## 🏃 Run

\# Test individual exercises:

python squat\_checker.py

python pushup\_checker.py

python plank\_checker.py

python lunge\_checker.py

python deadlift\_checker.py

Press 'q' to quit any exercise.

## 📊 Day-by-Day Progress

- **Day 1:** Basic pose detection + webcam setup ✅

- **Day 2:** 5 exercise analyzers with angle calculations ✅

- **Day 3:** Coming tomorrow - Bug fixes + improvements

## 🤝 Contributing

This is a learning project built in public!

Known issues are documented above. PRs welcome, especially for:

- Angle smoothing algorithms

- Better rep counting logic

- Additional exercises

## 📝 License

MIT

## 🔗 Building in Public

Follow the journey: [Your Twitter/X handle]

---

**Status:** Alpha v0.1 - Working prototype with known tracking issues

**Last Updated:** November 5, 2025

🔌 API Usage

Start the API Server

cd backend
python main.py

Server runs on: http://localhost:8000

API Endpoints

Health Check:

GET /api/health

List All Exercises:

GET /api/exercises

Analyze Form (Upload Video):

POST /api/analyze
- Form data: video (file)
- Form data: exercise (string)

Interactive API Docs

Visit http://localhost:8000/docs for interactive API documentation and testing!

Example API Response

{
  "success": true,
  "exercise": "squat",
  "form_score": 85,
  "rep_count": 10,
  "feedback": [
    "✅ Good depth!",
    "⚠️ Keep back straight"
  ],
  "message": "Analyzed 90 frames"
}