An AI-powered traffic surveillance system that detects multiple types of traffic violations using computer vision and deep learning models. Each violation type is handled by a separate, independently optimized model to allow flexibility, better benchmarking, and modular updates.

• 🎯 Helmet Detection – Detects riders without helmets using YOLOv8.
• 😷 Face Mask Detection – Identifies whether individuals are wearing masks (ResNet).
• 🧍 Triple Riding Detection – Detects more than 2 people on a bike (YOLO/ResNet).
• 🚫 Red Light Violation Detection – Detects vehicle movement beyond stop-line (OpenCV logic).
• 🔍 Modular Architecture – Each model runs independently; combine or deploy separately.

helmetdetection-using-yolo8m-opencv/
├── datafortesting-and-trainning/
├── testing/
├── train/
├── models/
├── scr/
├── main.py
├── dynamic signal handling.py
├── helmetvc.py
├── requirements.txt
└── README.md

1. Clone the repo

   git clone https://github.com/sairam-s0/Smart-Traffic-Violation-Detection-System.git
   cd Smart-Traffic-Violation-Detection-System

2. Install dependencies

   pip install -r requirements.txt

3. Download model weights (Place your YOLO and other model weights in the appropriate folders)

Each module runs independently:

# Helmet detection
cd Helmet-Detection
python helmet_detect.py --source your_video.mp4

# Mask detection
cd ../seatbelt-Detection
python mask_detect.py --source your_video.mp4

# Red light violation detection
cd ../Red-Light-Violation
python redlight_detect.py --source your_video.mp4

• 📈 All models show smooth learning curves with no signs of overfitting.
• ✅ License Plate model is the strongest — highly accurate and stable.
• 🚧 Seatbelt model performs well in general, but tighter localization (mAP@0.5:0.95) could be improved.

Instead of a single monolithic pipeline, this system treats each violation type as a separate module for:

• Easier maintenance and debugging
• Model-specific training & optimization
• Scalable deployment (microservices, edge devices)

Instead of a single monolithic pipeline, this system treats each violation type as a separate module for:

• Easier maintenance and debugging
• Model-specific training & optimization
• Scalable deployment (microservices, edge devices)

This project is licensed under the MIT License. See LICENSE for details.

Sairam GitHub: @sairam-s0

Pull requests, suggestions, and bug reports are welcome. Please open an issue or submit a PR with improvements or additional models.