Comparative Analysis of CNN And YOLOv5 For Automated Skin Burn Severity Classification

Abstract

The main objective of this study is to improve burn depth evaluation accuracy through automated deep learning models including Convolutional Neural Networks (CNNs) and YOLOv5. The standard medical approach for evaluating burns involves subjective assessments which generate unpredictable variations between observers thus resulting in incorrect diagnoses. We developed an automated system that provides immediate precise classification of burn wounds into first, second and third-degree categories. The research evaluates the performance of CNN-based models together with YOLOv5 by measuring their detection accuracy and processing speed and operational efficiency. Our research shows YOLOv5's single-stage detection system delivers better performance in accuracy and real-time detection compared to CNNs. The study demonstrates how automated burn wound evaluation can be enhanced for clinical practice which will decrease misdiagnosis rates while enhancing patient care.