Abstract:
Fashion plays a crucial role in self-expression and social interaction. This independent study proposes a virtual fitting room application that helps users select outfits based on celebrity fashion trends using computer vision and deep learning techniques. The system employs hybrid segmentation (semantic and instance segmentation) to accurately detect clothing items and classify their styles.To enhance image quality, Contrast Limited Adaptive Histogram Equalization (CLAHE) is used as a preprocessing step, improving model performance in detecting and analyzing clothing patterns. The system integrates Convolutional Neural Networks (CNNs), Plain Neural Network (NN), Support Vector Machines (SVM), and Random Forest into an ensemble model to achieve high accuracy in clothing classification.The final application provides users with a style similarity percentage comparing their outfit with a database of celebrity styles, helping users make informed fashion choices. Model performance is evaluated using cross-validation, accuracy metrics, and confusion matrices.The results indicate that the hybrid segmentation approach, combined with CLAHE-enhanced preprocessing and an ensemble learning framework, improves classification accuracy. This study contributes to the development of intelligent fashion recommendation systems and supports future applications in virtual fashion assistants and Metaverse integration.Keywords: Computer Vision, Fashion, Deep Learning, Virtual Fitting Room, CLAHE, Image Processing, Machine Learning.