• Main
  • Insights
  • The SmartMerch team created a neural network for retail powered by Google Cloud Platform

The SmartMerch team created a neural network for retail powered by Google Cloud Platform

Services & Solutions Cloud Solutions
Vendors Google Cloud

Now the team has the necessary technical basis to build up expertise and take leading positions in the domestic IT market in the FMCG-related development segment.

Background

The FMCG market, as a dynamically developing area, actively utilizes digital technologies. To improve business efficiency, the industry seeks to save data processing time by minimizing or eliminating manual labor. The SmartMerch project offers digital merchandising solutions. This area combines automation technologies for the product range audit at the points of sale using photography. As a result, clients receive a tool for the most objective analysis of a large array of photographs of products in various packaging. Having conflated their experience, the project team began to use neural networks and machine learning technologies in digital merchandising. This would enable image recognition with significantly higher accuracy and speed than a human being is capable of.

The technical implementation of the project required substantial computing capacity, and the team opted for Google Cloud Platform. SmartMerch appreciated the possibility to launch a project without building an IT infrastructure from scratch and the flexibility of resource consumption depending on business goals. The platform has a clear administration mechanism. It makes it easier for developers to learn the infrastructure.

Solution

“The project of migration from a local server to the cloud took about a month. The deployed solution is administered by two in-house SmartMerch specialists. With migration to the Google Cloud Platform, TPU – a Google service for developers who want to run machine learning processes in the cloud – is available to the specialists. The SmartMerch team uses Compute Engine, Cloud Storage, GPU, DataLab, StackDriver. In the future, we are considering connecting CloudSQL and G Suite," says Nikita Kostromov, SmartMerch Technical Manager.

One of the advantages of working with the Google Cloud Platform is convenient payment for consumed capacities. In organizing this process, SmartMerch was assisted by Softline, the only Russian holder of the Google Cloud Premier Partner status. As the vendor's partner, Softline gave the client an option to make monthly non-cash payments in rubles. In addition, consultations with Google engineers on cloud migration and technical support of the solution were organized for SmartMerch.

Results

“Now there are several domestic and foreign solutions on the market, but we are the first Russian company that was able to provide quality comparable to Western colleagues at a price attractive for customers. When creating the system, we achieved impressive results: the processing time is 2 seconds per image with an accuracy of 95%. Thanks to migration to Google Cloud Platform, we have received a powerful and resilient infrastructure to implement digital merchandising projects. Before that, we had several successful pilot launches. Now the team has the necessary technical basis to build up expertise and take leading positions in the domestic IT market in the FMCG-related development segment," says Maxim Arkhipenkov, creator and leader of the SmartMerch project.

“Softline actively promotes innovative Google Cloud solutions and helps companies migrate to the cloud platform, because it is a flexible, powerful and cost-effective solution. Our partnership with the vendor is aimed at creating the most comfortable conditions for the implementation of high-tech projects around the world: at any stage of work, customers have access to consultations with specialists on the implementation, maintenance and support of Google products," says Alexander Ivlev, Head of the Google Cloud Solutions Department at Softline.

See also similar solutions in this industry

Review our customers’ stories
Choose language