Поиск мебели по фото
Предметы интерьера
DesignScan лого

Search for similar images

System that finds interior items from a photograph

DesignScan лого

Main information

Client
DesignScan —
online furniture store
online store users
designers
Who is the service for?

Interesting fact

This is our first project with a company from the Baltics.

Problems that the customer came with:

Searching for a product can take up to twenty steps.
Some products cannot be found by text description.
Product descriptions can be subjective.
the client leaves
As a result

What the system gives:

helps to find objects with complex geometry
shortens the path from customer to product
For example, a person
liked a table in a cafe, and hewants the same home - you can take apicture of it and upload it to the search.

Development process

Given: a database of over a million images of interior items
We taught the neural network to remember:
- shape
- color
- texture
- context
The system receives photo at the entrance
Белое кресло
Returns the most relevant images, plus an indication of the category
Additional function
The service may offer interior items of a different category that match the style of the desired one.
Лампы
took
months
4
Principle of operation
User photographs an object
The photo is uploaded to the site
The search engine finds a similar product in the store catalog
Настольная лампа
search
Схожие настольные лампы
search
Схожие стулья
Комод
search

Technologies

EORA Mage

To develop this service, we used
Search engine for similar images from EORA.

Projects with EORA MAGE

KazanExpress

System that allows marketplace customers to search for products by photo.
KazanExpress лого

INTELSONLINE

System that allows you to quickly check a logo for plagiarism.
ИНТЭЛС лого

Difficulties of the project

Lack of open algorithms for visual search based on neural networks.
The need to formulate similarity criteria for testing the system.
Large volume of product categories.

Project team

International transaction support
Adaptation of EORA MAGE to new product categories
DesignScan лого
Manager
Computer Vision Engineer
Konstantin Kubrak
Viacheslav Schultz

Our plans

Create a mechanism for automatic retraining of the neural network for new categories of goods.
Contact us