Indonesia Facial Skin Disease Detection on Android Application Using TensorFlow Lite and Convolutional Neural Network

Masduki Zakarijah, Dimas Prasetyo, Dedi WijayaNon-Scopus10 Juli 2025Afiliasi: ID

Abstrak

This study focuses on the development of an Android application that utilizes a Convolutional Neural Network (CNN) with the MobileNet V2 architecture and transfer learning methods to detect facial skin diseases. The development process includes requirement specification, planning, modeling, construction, implementation, testing, and maintenance phases. The dataset consists of 700 images categorized into 7 disease classes, with 80% used for training and 20% for validation. Data were collected through direct observation to evaluate the application's accuracy and detection speed. The results show that the application achieved a detection accuracy of 83% and an average detection time of 0.5 seconds, using TensorFlow Lite on the Android platform. Data analysis was conducted using descriptive statistics to assess product design, accuracy, and detection speed.

Identifier

Scroll to Top