OLX Group operates a network of online trading platforms in over 40 countries under market-leading brands such as Avito, dubizzle, letgo, OLX, Stradia, Storia, and more, that are used by over 300 million people every month to buy and sell almost anything, creating win-win exchanges for people, their communities and the environment. OLX Group is a part of Naspers. Founded in 1915, Naspers is a global internet and entertainment group and one of the largest technology investors in the world. Operating in more than 120 countries and markets with long-term growth potential, Naspers builds leading companies that empower people and enrich communities. It runs some of the world’s leading platforms in internet, video entertainment, and media.
OLX has hundreds of millions active users worldwide. Some sellers try improving visibility of their items by creating duplicate listings with photos taken from different angles. We are researching deep learning techniques to automatically spot duplicate listings.
The goal of the project is to develop a model that can detect whether sets of images from two different listings show the same item(s). The main challenge of the work is to define and train a similarity metric that quantifies a “distance” between every pair of image sets, thus inducing a measure of similarity. Generally, metrics can be categorized into two kinds: pre-define metrics (i.e. metrics which can be fully specified without the knowledge of data) and learned metrics (i.e. metrics which can only be defined with the knowledge of the data). The latter one can be learned with (supervised) or without (unsupervised) labeled data. In this project, you will focus on supervised metric learning. An additional challenge is that image comparison shall not only look at the similarity between salient object(s), but also look at other objects and backgrounds of the images.