Paper Title
Personalized Outfit Recommendation System With virtual Try on
Abstract
In today’s fast-paced world, people tend to prefer shopping quickly and in a more personalized manner. In addition to this, the fashion industry has been proliferating in recent times. The project work comes up with a solution that recommends personalized outfits to the customer in a digitalized manner along with a Virtual Try-On model. Firstly, Real-time data capture seeks to collect data to analyze images and video footage in real-time. Many models can be used for this purpose, but one popular model is YOLO v5s.It was first introduced as an object detection model that integrated bounding box prediction and object classification into a single, end-to-end differentiable network. And then above image is sent to the Outfit segmentation module which uses U2-Net for salient object detection (SOD). This part is followed by a pre-built Deep Face model for Age and Gender Classification. In addition to this, the Pre-trained Resnet-50 architecture is used to create the embedding’s of the inventory data and input data. These data are used to produce the desired amount of recommendations based on the K-NN algorithm. Lastly, ACGPN is an AI-powered system that uses computer vision and machine learning algorithms to generate a realistic Virtual Try-On experience for customers provides the users with a comfortable way to check the outfits which they would like to try on and recommend the top-rated outfits-based present in the showroom as well according to what they wear. The accuracy of the Category is 0.94 and the article type is 0.83. This not only minimizes the decision-making confusion amongst people but also provides them with aVirtualTry-On.
Keywords-Recommendation System, Virtual Try-On, DeepLearning,Object detectionModel,Machine Learning.