Applications of Machine Learning to the Multiclassification of Images

Abstract

Classification of clothing finds wide application in fashion industry, social media, e-commerce, forensics, criminal law etc. and thus requires improved accuracy in classifying them. In this project we review Machine Learning (ML) concepts and work on various methods of Convoluted Neural Networks. We apply them to a data set of images to create a model that classifies the images as different clothing items. The Hyper-parameter Lambda is used in a Regularization function. The model loss reduces with the addition of the Hyper-parameter to the model. However, upon increasing the value of Lambda there is no significant improvement on the model. Hence, we can conclude that Lambda = 0.0 is the best model with minimum loss.