Virtual Try-On Systems: A Deep Learning Approach to Real-Time and Image-Based Garment Fitting

Résumé fourni par la source

The retail industry has changed with the rise of online garment shopping, but there are still many challenges. Some of the common issues include improper sizing and fit, which causes a high rate of garment returns and customer dissatisfaction. Traditional online shopping uses static images and sizing charts, which most of the time fail to represent the actual fitting of garments onIndividual Bodies. This paper describes the virtual try-on system in the quest to revolutionize the online shopping experience, which is interactive and personal. The user must submit simple front-view images of themselves, which are further analyzed by the ResNet model, U2Net, and OpenCV to produce the clothing mask. This model correctly measures the body of the person taking pictures and offers aprecise and personalized fit. Customers can preview themselves in various garments and customize options such as color and style through a web interface. The system also allows users to takekey measurements for creating made-to-measure garments. This method ensures quick feedback through the real-time processing capabilities of a pre-calculated garment database. The training accuracy obtained by the Resnet model is 85%. Overall, the proposed work enhances fit accuracy, reduces return rates, and improves customer satisfaction, representing a significant advancement in online retail for both consumers and manufacturers garments.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.