Helmet Detection System

This project detects motorcycle riders without helmets using a custom-trained YOLOv8 model. When a violation is detected, the system saves:

• The entire bounding box region of the rider (not just the helmet or head)
• The entire bounding box region of the number plate
• Both are saved in a timestamped folder for easy tracking and record-keeping

---

Features

• Detects riders with and without helmets
• Identifies and saves number plates of violators
• Generates a CSV file with paths to each violation image (optional)
• Organized folder structure by date and time

---

Workflow

Step 1: Dataset

We used a custom dataset from Kaggle:

Rider with Helmet / Without Helmet / Number Plate Dataset

---

Step 2: Training

The training logic is implemented in train_model.py, which uses YOLOv8 to train on the custom dataset.

---

Step 3: Inference

Inference is handled by the rider_safety_check.py script.

What the script does:

• Loads the trained YOLOv8 model
• Applies it to a video to detect:
  • Riders
  • Helmet status (with helmet / without helmet)
  • Number plates
• Draws bounding boxes on the video
• Saves the full bounding box region (not cropped) of:
  • The rider (without a helmet)
  • The number plate
• Stores each detection in a folder named using the exact timestamp:
  track/YYYY-MM-DD_HH-MM-SS
• Automatically stops after saving 5 valid violations to avoid redundancy

---

Step 4: CSV Records

A CSV file named records.csv is generated containing:

rider_paths	number_plate_paths
track/...	track/...

This file stores only the file paths of the detected rider and number plate images.
It is useful for referencing, automation, or further data processing.

---

Contents of track.zip

• Output video showing detected violations
• Rider images (full bounding box)
• Number plate images (full bounding box)
• records.csv with file paths of saved images

---

Video Credits

The video used for inference was sourced from Pexels.
All credit goes to the original video creators on Pexels for providing high-quality content for testing and demonstration purposes.

---

Requirements

Install the required Python packages:

pip install -r requirements.txt