Implementasi Transfer Learning Menggunakan Convolutional Neural Network untuk Deteksi Jenis Kulit Wajah

In Indonesia, extreme tropical climate conditions with high humidity and sun exposure increase the risk of facial skin problems for the community. Facial skin that is not properly cared for is often prone to disorders, ranging from dry skin, oily skin, to acne. However, Indonesian people's awareness of the importance of maintaining healthy skin is still relatively low, which is exacerbated by limited time and access to consult a dermatologist. Most people may not know their skin type, even though each skin type requires different care to stay healthy and avoid more serious skin problems. To answer this problem, this study aims to develop an iOS-based application that is able to automatically detect facial skin types using transfer learning with a Convolutional Neural Network (CNN) architecture. The model was developed by training a dataset of facial images to classify skin types such as dry, oily, normal, and acne-prone, and integrated into an iOS application for real-time analysis through user facial images. The evaluation results showed a model accuracy of 87% and an application accuracy of 83.3% in identifying facial skin types. It is hoped that this application will help Indonesian people better understand their skin conditions and obtain appropriate treatment recommendations to maintain healthy skin in a tropical climate.