Bangladeshi Online Market Product Classification Using Deep Learning

Abstract

Artificial Intelligence, is becoming the most powerful sector current days, especially deep learning or machine learning where program started to learn from our world’s data or experience. In future we can have an assistant in our home who can do online shopping for us, can bring the product what we tell it to bring. Online market is a vast network of current world where maximum people want to buy product without going to shopping mall. In Bangladesh approximately 2,000 e-commerce sites and 50,000 Facebook-based outlets delivering almost 30,000 products a day. It will rise day by day. In online market there are different kind of product with different categories. In this paper, I have proposed the convolutional neural network (CNN) based approach for classifying product of five different categories product with thirty tree types of subcategorical product from Bangladeshi popular online market from 1725 product images. For better research, I have used a model of CNN and identify a product with higher accuracy. Convolutional Neural Network (CNN) calculations, a machine learning system extensively applied to train program and identify product with class and subclass from given a known type picture. The outcomes model of this project can be applied to future robots or intelligent program or any kind of online ecommerce site’s admin panel to detect real product identical with given title. For better outcomes and better accuracy, I used RGB color model of the picture and trained the network with powerful Nvidia GPU. The prepared model accomplished an exactness of 95.14% on test set, showing the achievability of this methodology.