Abstract:
In this era of rapid advancement in technology, significance of better virtual shopping experience has increased greatly. Customers now-a-days prefer to save time by looking for something specific rather than wasting time in impulsive buying.
In order to cater this problem, we present a solution for the customers that reduces the search time by letting them snap a photo of their favorite outfit and search the brands’ catalogues based on the that picture. This saves the precious time of the customer that is usually spent going through every piece in the catalogue and searching particularly for the outfit that he/she wishes to purchase.
Our system is developed using deep learning approach. Currently there are a number of people working on solving the image retrieval problem using deep learning. A number of these are concentrated towards supervised learning of Convolution Neural Network. We kicked it up a notch by implementing the solution via unsupervised learning of Convolution Neural Network. There are a few advantages of using unsupervised learning over supervised one, for instance, there is no overhead of manually annotating the data for training purpose but at the same time evaluating the results of unsupervised learning and finding the best similarity criteria for the process is a challenge. Nevertheless, a lot of effort is being put to explore this domain as it has the potential to fundamentally transform the field of machine learning.
