Deep learning model for Cutaneous leishmaniasis detection and classification using Yolov5

Abstract

Scars and skin ulcers are a feature of many skin diseases, including cutaneous leishmaniasis (CL). In endemic areas and developing countries, infection identification remains a challenge for the physician, the health system, and the patient. This study introduces an important new diagnostic method for rapid detection and accurate diagnosis by establishing a YOLOv5 training model to recognize cutaneous leishmaniasis (CL), under normal conditions based on the YOLOv5 network depends upon a deep learning approach. The development environment is related to Python language. The training data set contains a total of 160 photos taken from a mobile phone camera and converted to grayscale images to extract characteristic features and then applied image processing techniques such as flipping, rotating, and resizing to increase the dataset information. Image labeling was identified with a dermatologist to ensure the injury of cutaneous leishmaniasis (CL).In our approach, images were classified using polygonal bounding boxes to identify areas of interest so that dataset was divided into validation, training, and testing. flowed by feeding the dataset in the YOLOv5s. Our model was able to locate cutaneous leishmaniasis (CL) and achieved high accuracy in detecting and classifying infections, with an average accuracy of 70%. Which provides a speed technology for detection.