Razvoj mobilne aplikacije za odkrivanje kožnih bolezni s slik

Vidic, Eva (author), Klemenc, Bojan (mentor)

Problem tradicionalnega diagnosticiranja kožnih bolezni je, da je postopek zamuden in včasih nedostopen pacientom. V diplomski nalogi razvijemo mobilno aplikacijo za zaznavanje kožnih bolezni s slik, ki uporablja konvolucijsko nevronsko mrežo (CNN). Za učenje modela s prenosom znanja smo uporabili podatkovno zbirko, ki je javno dostopna na portalu Kaggle in že naučen model ResNet-50. Aplikacija omogoča hitro klasifikacijo različnih kožnih bolezni, kar uporabnikom omogoča pridobivanje informacij o njihovem zdravstvenem stanju in spodbuja zgodnje ukrepanje. Model, ki smo ga razvili, je na testni množici dosegel 84 % točnost.

• Language: Slovenian
• Keywords: kožne bolezni, klasifikacija, učenje s prenosom znanja, konvolucijske nevronske mreže (CNN), mobilna aplikacija
• Work type: Bachelor thesis/paper
• Typology: 2.11 - Undergraduate Thesis
• Organization: FRI - Faculty of Computer and Information Science
• Year: 2025
• PID: 20.500.12556/RUL-173300
• COBISS.SI-ID: 253653507
• Publication date in RUL: 15.09.2025

Secondary language

• Language: English
• Title: Development of a mobile application for skin disease detection from images

Abstract

The problem with traditional diagnosis of skin conditions is that it is time-consuming and sometimes inaccessible to patients. In this thesis, we develop a mobile application for detecting skin conditions from images using a convolutional neural network (CNN). We trained and fine-tuned a ResNet-50 model on a publicly available dataset from Kaggle. The application enables rapid classification of various skin conditions, allowing users to obtain indicative information about their health and encouraging early intervention. The developed model achieved an accuracy of 84 % on the test set.

• Keywords: skin diseases, classification, transfer learning, convolutional neural networks (CNN), mobile application