🚨 Road Safety Assistant - Helmet Violation Detection System

A real-time computer vision system to detect helmet violations using YOLOv8 and OpenCV, with an interactive Streamlit dashboard. This project helps improve road safety by automatically identifying motorcyclists riding without helmets from traffic surveillance videos.

---

📌 Features

• ✅ Real-time helmet detection using YOLOv8
• 🚷 Logs and saves cropped images of helmet violators
• 🧠 Avoids duplicate entries using object tracking
• 📈 Interactive Streamlit dashboard to:
  • Upload traffic videos
  • View logged violations
  • See cropped snapshots
  • Export logs as CSV

---

📸 Demo

---

🚀 Getting Started

1. Clone the Repository

git clone https://github.com/your-username/road-safety-assistant.git
cd road-safety-assistant

2. Install Requirements

pip install -r requirements.txt

3. Download YOLOv8 Helmet Model

Place the YOLOv8 model (trained for helmet/no-helmet detection) in the models/ directory.

Recommended:

• helmet_yolov8.pt (or your own trained model)

models/
└── helmet_yolov8.pt

4. Run the Streamlit App

streamlit run app.py

📁 Project Structure

road_safety_assistant/
├── app.py                     # Streamlit UI
├── detectors/
│   └── yolo_detector.py       # YOLOv8 detection + tracking
├── utils/
│   └── ui_utils.py            # Video processing, logging
├── data/
│   └── test_videos/           # Input videos
├── logs/
│   ├── frames/                # Cropped violation images
│   └── violations.csv         # Log of all violations
├── models/
│   └── helmet_yolov8.pt       # YOLOv8 helmet detection model
├── requirements.txt
└── README.md

📊 Output

📁 logs/violations.csv:

timestamp,track_id,class,confidence,image
2025-07-06 17:12:34,2,Without Helmet,0.89,logs/frames/violation_2025-07-06_17-12-34.jpg

📁 logs/frames/:contains all cropped images of violators.

🧠 Model Information

The model is trained on a helmet detection dataset with two classes:

• With Helmet
• Without Helmet