README.md

Triple-Rider Detection

Project Overview

Triple-Rider Detection is a computer vision project that detects motorcycles with more than two riders and checks for helmet compliance. Additionally, it identifies if the motorcycle is being used on a mobile device. This project utilizes YOLOv8, trained on a custom dataset of over 6,000 images, to achieve high-performance detection.

Table of Contents

• Project Description
• Installation Instructions
• Usage Instructions
• Dataset Information
• Training Details
• Preprocessing
• Sample Output Images

Project Description

Triple-Rider Detection aims to enhance motorcycle safety by identifying whether a motorcycle has more than two riders and ensuring all riders are wearing helmets. The project also detects if the motorcycle is being used on a mobile device. The system uses YOLOv8 for object detection, providing real-time feedback on helmet compliance and usage status.

Installation Instructions

1. Clone the Repository:

   git clone https://github.com/yourusername/triple-rider-detection.git
   cd triple-rider-detection

2. Set Up the Environment:

   Create a virtual environment and install dependencies:

   python -m venv env
   source env/bin/activate  # On Windows use: env\Scripts\activate
   pip install -r requirements.txt

3. Install Roboflow Inference SDK:

   pip install inference-sdk

Usage Instructions

Using the Roboflow Inference SDK

1. Obtain a Roboflow API Key:

   To use the Roboflow API, you need your own API key. Follow these steps:

   • Sign up for a Roboflow account at Roboflow.
   • Create a new project and upload your dataset.
   • Retrieve your API key from the Roboflow dashboard.

2. Update API Key and Model ID:

   Replace the YOUR_API_KEY, YOUR_IMAGE.jpg, and model_id with your actual values in the code snippet below.

   # import the inference-sdk
   from inference_sdk import InferenceHTTPClient
   
   # initialize the client
   CLIENT = InferenceHTTPClient(
       api_url="https://detect.roboflow.com",
       api_key="YOUR_API_KEY"  # Replace YOUR_API_KEY with your actual API key
   )
   
   # infer on a local image
   result = CLIENT.infer("path/to/your/image.jpg", model_id="3riders/2")

3. Run Inference:

   Save the above code to a Python file and execute it to perform inference on your image.

Dataset Information

The dataset used for training includes a custom collection of images created specifically for this project. It is divided as follows:

• Train Set: 87% (5,295 images)
• Validation Set: 8% (514 images)
• Test Set: 4% (248 images)

This custom dataset was designed to provide a diverse range of scenarios to improve the accuracy and robustness of the model in detecting motorcycles with more than two riders, helmet compliance, and mobile usage.

Training Details

• Model: YOLOv8
• Training Set Size: 6,000 images
• Performance Metrics:
  • mAP: 81.7%
  • Precision: 80.8%
  • Recall: 75.0%

Training Graphs

Preprocessing

• Auto-Orient: Applied
• Resize: Stretch to 640x640
• Augmentations:
  • Flip: Horizontal, Vertical
  • Rotation: 90° Clockwise, Counter-Clockwise
  • Crop: 0% Minimum Zoom, 22% Maximum Zoom
  • Rotation: Between -11° and +11°
  • Shear: ±12° Horizontal, ±12° Vertical

Confusion Matrix

Sample Output Images

Image 1

Description: Example of motorcycle with more than two riders detected.

Image 2

Description: Example of helmet compliance detection.

Image 3

Description: Example of detection results with bounding boxes and labels.

Image 4

Description: Example of motorcycle detection with mobile usage.