Motorcycle Helmet Detection Model

The Motorcycle Helmet Detection Model is a computer vision system developed to automate motorcycle helmet compliance detection from road images and videos. It combines YOLOv8 object detection with spatial relationship analysis to transform raw object detections into rider-level helmet compliance decisions.

Manual helmet compliance monitoring is labour-intensive and difficult to scale across large road networks. Key issues include:

• Time-consuming manual roadside enforcement
• Difficulty monitoring high-volume traffic continuously
• Object detection alone cannot determine rider helmet compliance
• Real-world conditions such as occlusion, dense traffic, and low-light environments reduce detection reliability

The system combines deep learning object detection with rule-based spatial analysis to transform raw object detections into helmet compliance decisions. Major capabilities include:

• Multi-class Object Detection (Helmet, Human, Motorcycle)
• Rider Identification & Association
• Helmet Compliance Classification
• Rule-based Spatial Analysis
• Interactive Flask Dashboard
• Image Processing
• Performance Evaluation & Model Comparison

• Computer Vision Engineer
• Collected and curated external testing datasets covering multiple road scenarios
• Performed data cleaning, annotation validation, and exploratory dataset analysis
• Participated in training and evaluating eight YOLOv8 model configurations
• Compared model performance using Precision, Recall, mAP@50, mAP@50–95, and confusion matrices

• Computer Vision
• Deep Learning
• Object Detection
• YOLOv8
• Dataset Engineering
• Data Cleaning & Annotation
• Model Training & Evaluation
• Rule-based AI
• Flask Development
• Performance Analysis
• Image & Video Processing
• Team Collaboration