A Mobile Application Using MobileNetV2 for Classification of Skin Diseases

Abstract

This study presents the development of a mobile application for classifying skin conditions into Melanoma, Acne, and Healthy skin. A dataset of 214 Melanoma images from Dermnet, 395 Acne images, 400 Healthy skin images, and 101 Melanoma images from the RAW10000 database was used. Pre-processing techniques such as rescaling, sharpening, bilateral deceleration, and min-max normalization were applied. Feature extraction was performed using HOG, VGG16, and MobileNetV2, followed by classification with SVM, RF, and ANN models. MobileNetV2 achieved the highest accuracy at 94.30%. The backend was developed with Django and MySQL, and the user interface with Figma and React. The project demonstrates significant accuracy and practical application potential, enhancing mobile device access to dermatological diagnostics and early diagnosis.