Image similarity-based fashion recommendation web application using angular and machine learning

Abstract

People utilize fashion as a significant form of self-expression for a variety of reasons. It appears to be an essential component of every person’s existence in contemporary civilizations, from routine activities to noteworthy moments and events. Since there is a great demand for fashionable goods, the fashion sector is viewed as desirable and lucrative. Although there is a great chance for businesses to engage in industries related to fashion because of the enormous demand, there are also a number of dif ficulties in meeting the demands of the market. Systems that recommend clothing have been developed to meet these needs. The complex conceptions of this domain and their relevance have been developed, justifying fashion domain-specific traits. Retrieving clothing items, recommending complementary items, outfit recommen dations, and capsule wardrobes are the four core functions of image-based fashion recommendation systems. There have been three primary eras and the most recent breakthroughs depicted in an evolutionary trajectory of image-based fashion rec ommend systems with regard to computer vision advancements. In this project, a CNN-based transfer learning approach for recommending fashion items was imple mented. And for visual representation, an angular template was used to show the results. For implementing this, I tried various transfer learning image classification algorithms, among which Resnet50 got the best result. Also, two classifiers were used to improve the performance of the algorithm.