Getrealprice began using neural networks to create new clothing designs and predict their sales

Russian company Getrealprice, which provides a service to monitor prices and assortment on the Internet, with the help of neural networks has created directories of "machine design" in the categories of "clothing" and "jewelry" (rings and earrings).

"We have actively embraced new technologies that were previously inaccessible without GPU devices with a large memory capacity. Everyone can order catalogues of "machine design" products in the "clothing" category, as well as in two jewelry categories - "earrings" and "rings", the catalogue of which has a total of about four thousand products," said Nikita Mikhailov, CEO of the company.

The designs presented in the catalogs are created by generative neural networks, which, in turn, are trained by photos of products with high sales.

According to Mikhailov, such catalogues may in the future become a working tool for designers when designing products.

Predicting the sale of a product from a sketch

In addition, Getrealprice has also developed a service for retailers that is able to predict the sales of a garment based on its sketch. Using SketchFinder neural network, it's possible to find similar products already on the market from a sketch or a photo of a garment and show their real sales volume in a fixed period.

SketchFinder works with the following clothing categories: pants and shorts, outerwear, turtlenecks, jumpers and cardigans, jeans, vests, jumpsuits and dungarees, suits, longsleeves, jackets and coats, shirts, sweatshirts, sweatshirts, t-shirts and tank tops, hoodies, clothing accessories, as well as bags, boots, sneakers, sneakers, moccasins, slip-ons, shoes and flip-flops.

The SketchFinder neural network monitors in three stages:

1. The neural network processes a sketch, a sketch or a photo of an item of clothing. And the sketch can be drawn by hand, on a computer, or using a ready-made vendor model. The more details there are in the sketch, the better will be the selection of similar items.

2. Then the SketchFinder neural network divides the drawing into elementary parts ("vectors"), defines the patterns to be used.

3. After the information processing the neuronet produces the top 30 similar existing items with their sales history in items and with the current market price, and an instant selection of analogues among all the marketplaces. Search for similar style items can be performed both within the same category, and among the items of other categories.

About Company

Getrealprice is a service for monitoring prices and assortment on the internet with further processing of the obtained data in its software in order to set the optimal daily price for each product in any online store.

The company is among the top five companies in its sector in the Russian Federation, with Getrealprice being the leader in Moscow. For the last 3 years, the company's sales have grown by more than 20%, while its domestic market share is more than 30%.

Getrealprice works both with Russian companies to support competitive operations in the local market, and with foreign companies to facilitate the rapid operational expansion of the range of global exporters to other markets.

Products

The company constantly releases new products and solutions, which are developed based on the business needs in today's realities. Getrealprice services significantly differ from proposals of consulting companies and system integrators in the fact that they are autonomous, easy to implement and error free. Getrealprice's service is easily managed by a mid-level e-commerce employee. In addition, the company trains its product - both independently and through partner business schools.

The service has a number of areas (Diy and Household, Fashion, Ce, Beauty & Parfum, Sport), for each of which a set of offers has been developed:

• Manufacturers: monitor resellers' compliance with the recommended retail price;
• retailers: monitor their prices along with the prices of their competitors;
• Marketplaces: monitor their own and their competitors' prices and set quoted prices automatically.

Technologies

The Company uses a number of technologies to implement its activities:

• proprietary registered software and Databases (DB) to compare prices of goods in online stores;
• Traffic management: management of request traffic, database load, conversion rates and queries;
• human behavior emulation: robots that process prices are able to adapt to human behavior in terms of listing algorithm and content loading;
• has its own physical farms for parsing in different countries, being a Russian alternative to Crawlera, GeoSurf, and Luminati;
• applies its own tool for product classification of a large and fragmented assortment cluster, which allows managing any commodity nomenclature, including specific ones (pharmaceuticals, automotive spare parts);
• GAN and CNN neural networks are actively used, in particular to create new product designs or, for example, to emulate human behavior (Skectfinder products, Gan-designs, GetRealPosition).