helmet-detection-yolov8

Model Information

This repository contains the YOLOv8n model used by ViolaWatch for helmet detection.

Model Details

• Architecture: YOLOv8n
• Task: Object Detection
• Classes:
  • With Helmet
  • Without Helmet
• Framework: Ultralytics
• Model format: PyTorch (.pt)
• Best mAP50-95: 0.32925

Using the Model

Install Ultralytics:

pip install ultralytics

# 🚦 ViolaWatch v2 — Traffic Violation Detection System

Real-time detection of **no-helmet motorcyclists** and **seatbelt violations**,  
with **license plate recognition**, MySQL storage, Desktop GUI, and web dashboard.

---

## Quick Start

### Desktop GUI (Recommended for local use)
```bash
pip install -r requirements.txt
python main.py

Web Server (Local or hosted)

pip install -r requirements.txt
python main.py --web
# Open http://localhost:5000

MySQL Setup

Option A: Local MySQL

CREATE DATABASE violawatch;
CREATE USER 'viola'@'localhost' IDENTIFIED BY 'your_password';
GRANT ALL ON violawatch.* TO 'viola'@'localhost';

Edit config/config.ini:

[database]
host = localhost
user = viola
password = your_password
database = violawatch

No MySQL? The app auto-falls back to SQLite — just run it!

Option B: Free MySQL cloud (for hosting)

• PlanetScale — free 5GB MySQL (https://planetscale.com)
• Railway MySQL — free tier (https://railway.app)
• Aiven — free 5GB (https://aiven.io)

Set environment variables instead of config.ini:

DB_HOST=...  DB_USER=...  DB_PASSWORD=...  DB_NAME=violawatch

Free Hosting Deployment

🚂 Railway (Easiest — recommended)

1. Push code to GitHub
2. Go to https://railway.app → New Project → Deploy from GitHub
3. Add MySQL plugin → copy connection vars to environment
4. It auto-detects Procfile and deploys

🎨 Render

1. Push to GitHub
2. https://render.com → New Web Service → connect repo
3. Render reads render.yaml automatically
4. Add MySQL env vars from Aiven/PlanetScale

🐳 Docker (anywhere)

docker build -t violawatch .
docker run -p 5000:5000 \
  -e DB_HOST=... -e DB_USER=... -e DB_PASSWORD=... \
  violawatch

☁️ Fly.io

fly launch --name violawatch
fly secrets set DB_HOST=... DB_USER=... DB_PASSWORD=...
fly deploy

Video Upload (Testing)

From the web dashboard or Desktop GUI:

1. Click Upload Video File
2. Select MP4/AVI/MOV file (up to 500MB)
3. Processing runs in background — live preview updates in real-time
4. Violations auto-saved to database with snapshots

Supported formats: mp4 avi mov mkv wmv webm

Camera Sources

Detection Accuracy

*Upgrade to custom-trained YOLOv8 model for 90-95%+ accuracy:

# Download free datasets from: https://universe.roboflow.com
# Then in config/config.ini:
[models]
helmet_model = models/helmet_yolov8m.pt
seatbelt_model = models/seatbelt_yolov8m.pt

Project Structure

violawatch/
├── main.py              ← Entry point (GUI or web)
├── server.py            ← Flask web API
├── Procfile             ← Railway/Heroku deploy
├── Dockerfile           ← Docker deploy
├── render.yaml          ← Render deploy
├── railway.toml         ← Railway config
├── requirements.txt
├── config/
│   ├── config.ini       ← All settings
│   └── __init__.py      ← Config loader
├── core/
│   └── detector.py      ← Detection engine
├── utils/
│   ├── frame_processor.py  ← YOLOv8 pipeline
│   └── plate_reader.py     ← OCR
├── database/
│   └── db_manager.py    ← MySQL + SQLite
├── gui/
│   └── app.py           ← Desktop GUI (Tkinter)
└── web/
    └── index.html       ← Web dashboard

Environment Variables (for hosting)