Abstract:
Skin diseases plague a large segment of the global community and usually need a lot of care and personal attention. This project introduces a web-based Facial Skin Disease Detection and Skincare Recommendation System which uses deep learning and image processing algorithms to diagnose popular skin conditions and provide appropriate skincare recommendations. The system concentrates on the five main classes, which include Acne, Wrinkles, Vitiligo Disease, Rosacea, and Pigmentation, based on pictures in the Acne04 and Skin-Condition datasets. MobileNetV2, EfficientNetB3, and a custom Convolutional Neural Network (CNN) were trained and evaluated as three deep learning models. MobileNetV2 was chosen as the main model among them because it has a high accuracy and real-time image classification performance. The model works with images of faces in order to identify skin conditions and activate a customized recommendation engine of appropriate skincare products and ingredients. The app is developed using ReactJS (frontend) and Python FastAPI (backend) and a MySQL database to store user information, prediction history, and product suggestions. Being lightweight and strong enough, this architecture is cross-platform and can be extended in the future. This system seeks to fill the knowledge gap between dermatological visit and self-care by enabling users to have immediate skin assessment of the skin and make informed product decisions. It has the potential to be used in the management of personal skincare as well as supplementary to teledermatology service
