Modified Convolutional Neural Network for Fashion Classification

Abstract

Fashion classification is a domain which finds its applications in various fields like e-commerce platforms, social media and criminal identification with clothing similarity or dissimilarity. In this paper, we have used a modified version of convolutional neural network for classification and encompassing the identification of clothing items. Within the fashion classification category, we mainly concentrate on the multi-class classification of different types of apparels. The modified convolution neural network is applied on fashion classification data which reduces over-fitting. Here we have compared the accuracy of the CNN models and have achieved train accuracy and test accuracy of around 93% and 90% respectively which are better than previous works done by other researchers.