Design of Facial Skin Type Detection Application Using CNN with Inceptionv3 Model and Google Cloud Platform

Abstract

The advancement of Artificial Intelligence (AI) and Computer Vision technologies has significantly impacted the beauty industry, particularly in facial skin type detection. This study developed a mobile application that utilizes CNN with the InceptionV3 architecture deployed on the Google Cloud Platform (GCP). The system uses a dataset of 1,735 facial images categorized into normal, dry, oily, and acne-prone skin types. The photos were preprocessed and augmented before being processed by the CNN model. Firestore and Cloud Storage were used to maintain the data, while Cloud Run was used to publish the trained model into a Flask-based API. The accuracy, precision, recall, and F1-score reached 91.7%, 91%, 91%, and 91% respectively. Compared to previous studies, this system offers real-time classification through a lightweight mobile application integrated with cloud computing, aiming to improve accessibility and efficiency in dermatological analysis and personalized skincare services.