We also compare and report incremental gain achieved by our unified multi-task model over each individual task separately. This technique displays a better performance compared to the traditional triplet mining baseline, which uses single label/attribute information. Next, we introduce an offline triplet mining technique which utilizes information from multiple attributes to capture relative order within the data. Masking technique is used for designing the attribute classification. Our model consists of three different tasks: attribute classification, triplet ranking and variational autoencoder (VAE). We also highlight and address the challenges like image compression, cropping, scribbling on the image, etc, faced in reseller commerce. In this work, we design a visual search system for reseller commerce using a multi-task learning approach. Product attributes available in e-commerce have a great potential for building better visual search systems as they capture fine grained relations between data points. It allows resellers to accurately locate the products that they seek, unlike textual search which recommends products from head brands.

Visual search is of great assistance in reseller commerce, especially for non-tech savvy users with affinity towards regional languages.

Pinterest visual search