Transfer Learning with Deep Neural Networks for Image Classification in the E-commerce Industry

E-commerce has changed the way people shop and unleashed new levels of demand for goods by making it easier to buy and deliver goods to customers. There are now more choices for consumers with shorter production cycles thanks to fashion's digitization. Digital technology in the e-commerce industry is grabbing the attention of customers. The need for a system that facilitates product searching is therefore growing. Using fashion image classification, online retailers can quickly and efficiently manage the volume of their products. Use it to sort products based on their brands, types, or even styles. Fashion ensembles and even product recommendations could be created with it. Product filtering and search can benefit from the results of apparel image classification. For coarse-level image classification, a wide range of models have been successful, but only a few of them could be applied to fashion images because of the nature of the data: such images could belong to a single high-level topic like 'shirt’, ‘saree’, or ‘earrings’, but at the lower level be fine-grained (for example, 'cotton saree,' ‘long earrings’). The fine-grained hierarchy of each clothing store makes it difficult to collect a large dataset. For good predictive quality, research has shown that there is a need for machine learning techniques. Currently, there are new possibilities for vertical integration of research results into industrial or commercial applications due to recent advances in the image retrieval domain. One has to do with the difficulty sellers have when trying to upload images of their products to a marketplace and the ensuing manual tagging that occurs as a result. As a result of the classification errors, it is no longer listed in search results. Using an image-based search algorithm will allow customers to click on a picture of an object and find related products without having to type anything in. As a result, we have created a fashion industry-specific intelligent technique based on deep learning for quick searches and accurate product classifications by applying transfer learning and experiment results on proposed methodology proved that XCEPTION deep neural network model outperforms among all other networks.