Pages 399–412

Authors

Tinnapat Jaimunt and Varin Chouvatut

Abstract

This study presents a virtual fitting room application that recommends outfits based on celebrity fashion trends using computer vision and deep learning. The system applies hybrid segmentation (semantic and instance) for accurate clothing detection and uses CLAHE preprocessing to enhance image quality. An ensemble model combining CNN, plain NN, SVM, and Random Forest is used for style classification. The application provides users with a style similarity score compared to celebrity outfits. Evaluation through cross-validation and accuracy metrics shows improved performance, highlighting the potential of this approach for intelligent fashion recommendation systems and future Metaverse applications.