Cover Visuelles Produktmatching

Funding of blackbee’s visual product matching by the Free State of Saxony

Webdata Solutions has been awarded a substantial six-figure grant from the Sächsische Aufbaubank (SAB) for the technical development of the competitive intelligence suite blackbee. The SAB grant for the years 2020-2022 will be used to optimize blackbee’s latest technological feature: visual product matching. The project is implemented in collaboration with the University of Leipzig in the joint project “Visual Product Matching (VIP)”. The first kickoff meeting is scheduled for October 2019.

“We look forward to working closely with the University of Leipzig again to expand blackbee’s product matching capabilities and transfer them to new business areas,” said Carina Röllig, CEO of Webdata Solutions.

Goal of the collaborative project “Visual Product Matching”

blackbee‘s online market analysis is based on the accurate allocation of two data sets to the same product known as product matching. Using matched product pairs, blackbee compares the offers of several vendors. The biggest challenge in text-based product matching is the assignment of product descriptions. The same products are often described very differently, or the descriptions have a different level of detail. The goal of the “VIP” research project of Webdata Solutions in collaboration with the University of Leipzig is to develop algorithms that compensate for these limitations of text-based matching by including image information. To do this, the collaboration team must find answers to the question of how to compare images that, for example, show different perspectives of the same product, whose brightness and colour values differ or that are scaled differently. There are additional hurdles, such as in the fashion industry, where products are often presented as outfits. A fashion model seldom wears just one product; they usually wear more products together. In order to achieve an accurate matching, the market analysis software must also recognize the product in question in such a case.

Implementation and potential of ” Visual Product Matching (VIP)”

Together with the University of Leipzig, Webdata Solutions is developing an image recognition system that pre-processes product data with images and compares it with various methods. The system is based on an extensive data set provided by Webdata Solutions. As a result, the system outputs determined pairs or groups of identical products, which are then ready for further processing.

Generic image recognition has become a standard market offer. A number of companies, from large players to start-ups, are working on this topic. However, generic image recognition can only recognize people or objects, for example. In addition, there are providers whose applications can determine a product type. So far, however, there is no supplier who is able to recognize which specific product it is. This is where Webdata Solutions’ approach goes further by including the recognition of the actual product in the analysis. The collaboration team is thereby creating a unique system that will take online market analysis to a new level.

Don’t miss out on any information about blackbee and the project “Visual Product Matching” with our monthly newsletter. Subscribe now for free.