ABSTRACT
DEVELOPMENT OF CNN EFFICIENTNET AND RESNET ARCHITECTURE ON FACE SHAPE AND HAIR TYPE CLASSIFICATION FOR WEB-BASED HAIRSTYLE IDENTIFICATION SYSTEM
Name : Arafat Dhiya ’Ulhaq
NIM : 21051214024
Study Program : S-1 Information Systems
Department : Informatics Engineering
Faculty : Engineering
Institution : Universitas Negeri Surabyaa
Supervisor : I Kadek Dwi Nuryana, S.T., M.Kom.
Hairstyle selection is often subjective and not based on objective visual characteristics, thus reducing the level of personalization and user satisfaction. This research aims to develop a web-based hairstyle identification system capable of providing personalized recommendations based on visual data analysis. To achieve this goal, this research applies a deep learning approach by developing two Convolutional Neural Network (CNN) architectures on EfficientNetB0 for face shape classification task and ResNet50 for hair type classification.
The research method used is the Cross-Industry Standard Process for Data Mining (CRISP-DM), which includes the stages of business understanding, data understanding, data preparation with augmentation, modeling using transfer learning, evaluation, and implementation. The result of this research is a functional web application capable of classifying face shape and hair type from user-uploaded images. The system successfully integrates the two models to provide more accurate and personalized hairstyle identification, so it can be an objective solution in the digital beauty industry.
Keywords—Hairstyle Identification, Image Classification, EfficientNet, ResNet, Hair Personalization