Automated skin burn detection and severity classification using YOLO Convolutional Neural Network Pretrained Model

Skin burn classification and detection are one of topics worth discussing within the theme machine vision, as it can either be just a minor medical problem or life-threatening emergency. By being able to determine classify skin severity, help paramedics give more appropriate treatment for patient with different severity levels burn. This study aims approach this topic using computer vision concept that uses YOLO Algorithms Convolutional Neural Network models degree burnt area bounding boxes feature from these models. paper was made based on result experimentation dataset gathered Kaggle Roboflow, in which images labelled (i.e., first-degree, second-degree, third-degree). experiment shows comparison performance produced fine-tuned used similar algorithm implemented custom dataset, YOLOv5l model best performing experiment, reaching 73.2%, 79.7%, 79% before hyperparameter tuning 75.9%, 83.1%, 82.9% after F1-Score mAP at 0.5 0.5:0.95 respectively. Overall, how fine-tuning processes improve some effective doing task, whether by approach, selected real life situations.