Object Detection Application

A complete production-ready object detection application with ML training, tracking, Docker, FastAPI serving, CI/CD, and deployment.

Architecture

• Backend: FastAPI with YOLOv8 model for object detection
• Frontend: React app for uploading images and displaying results
• ML Tracking: MLflow for experiment tracking
• Containerization: Docker
• CI/CD: GitHub Actions
• Deployment: Render

Local Setup

Backend

1. Navigate to backend/ directory
2. Create venv: python -m venv venv
3. Activate: source venv/bin/activate
4. Install: pip install -r requirements.txt
5. Run training: python train.py
6. Start server: uvicorn main:app --reload

Frontend

1. Navigate to frontend/ directory
2. Install: npm install
3. Start: npm start

Docker

Build: docker build -t object-detection .

Run: docker run -p 8000:8000 object-detection

MLflow

View experiments: mlflow ui (from backend directory)

Deployment on Render

1. Connect your GitHub repository to Render
2. Create a new Web Service
3. Select the repository and branch
4. Set build command: (leave default or pip install -r backend/requirements.txt)
5. Set start command: uvicorn backend.main:app --host 0.0.0.0 --port $PORT
6. For frontend, deploy separately as Static Site on Render, build command: cd frontend && npm install && npm run build, publish directory: frontend/build

API

POST /detect - Upload image file, returns annotated image with bounding boxes