E-Commerce Product Image Classification using Transfer Learning
Bineet Kumar Jha, G G Sivasankari, K R Venugopal · 2021
E-commerce is the platform where it provisions the online businesses for selling and buying. Organizing and searching for products is a cumbersome process for service providers and customers. The considerable time are getting wasted in organizing and labeling the products. The most commonly used algorithm for image classification is the convolutional neural network (CNN). Training a huge dataset consumes a lot of computational resources and time. Image search is good when we don't have many details about the product. We have proposed a transfer learning approach based on visual geometry group-19 (VGG-19) and Inception V3 to overcome the issues related to classification, product identification, product suggestion, and image-based search. We have taken a Kaggle dataset of 70000, 28x28 labeled fashion images. The dataset is split into 85% training and 15% validation. The training and validation accuracy of the proposed transfer learning approach is better whereas the loss is lower as compared to CNN based approach.