RUS | ENG

Supported by:

Skoltech to develop neural network for retailers

The new solution is expected to help retail chains forecast sales and compute an ROI from marketing activity

Sep 26, 2016
The development of an image recognition system to take stock of goods in supermarkets is underway in Russia. Skoltech, the Skolkovo-based tech university, and Inspector Cloud, a private company, are responsible for project-related R&D in computer vision, portal Science & Technology RF reported, adding that the partners started work in August already.

According to market players, retail audit reports are seldom even 50% accurate, as stock-taking in supermarkets is still manual. The new solution is expected to help retail chains forecast sales and compute an ROI from marketing activity.

Inspector Cloud will be receiving photographs of shelves with food and other goods from a store to recognize the goods to be accounted for, and then generate a report with an estimated accuracy of 95%, using neural networks algorithms. Such approach has already been successfully tested in other areas, like recognizing people by analyzing photos and videos. With goods in supermarkets the task is a bit more complicated as each item has its typical shape, distinct characteristics, and a variety of tags.