Moscow

The solution allows a blind and deaf-blind user to easily and quickly understand, what banknote is in front of him. The solution is based on the artificial intelligence algorithms, that allow to instantly determine the denomination of a banknote. It is enough for a blind user just to bring the banknote to the phone from the camera side and its denomination will be announced by the application.

For deaf-blind people who neither hear nor see, the developers have introduced a special vibration mode - different types of vibrations correspond to different banknotes. The application can be installed on phones with Android or iOS operating systems. The project partner is “MegaFon”.

recognition

artificial_intelligence

Banknote identifier

Vocord

The solution is a complex for controlling traffic flow at crossroads and railway crossings. The device detects speeding violations, driving on road shoulder, driving on public transport lane and on the oncoming lane. The complex consists of the integrated VOCORD NetCam camera with fully-controllable lens, GLONASS/GPS receiver built into the camera and pulse infrared searchlight. The device can be complemented with a multipurpose radar speed meter installed on the body of the device. Data processing, numbers recognition and violations detection are conducted in the camera on the NVIDIA Jetson processor. The device does not require any external servers for operation. The main features of VOCORD MicroCyclops are low power consumption and a small range of requirements to data transmission channels.

Compact multi-lane complex for traffic control

GLONASS

VOCORD MicroCyclops

ABBYY

ABBYY FlexiCapture increases opportunities of working with data by applying Artificial Intelligence technologies. They allow to process any type of documents - both structured and unstructured documents (contracts, supplements, charters).
NLP (Natural language processing) identifies entities, roles and facts in the text. NER (Named Entity Recognition) technology allows to extract named entities, for example, full name, organization name or address. It is possible to extract any entities, such as life of a contract. Expanded NLP functionality allows to extract more complex concepts, including facts, roles and commitments, for example, fact of appointment to a post or contract conditions.
Due to machine learning and text processing using natural language, the platform automatically identifies, classifies, extracts, verifies and transfers data to information systems from any text or document used in business processes, such as client requests and transactions.

computer_vision

neural_networks

ABBYY FlexiCapture

Yandex

The infrastructure of the service is designed bearing high loads in mind to ensure that the system is available and fault-free even if the number of concurrent requests is high. Alice, Yandex Voice assistant, is based on SpeechKit.
The service processes audio and texts in four languages: Russian, English, Ukrainian and Turkish. Yandex SpeechKit composes speech from more than a million individual phonemes, the intonation is set by a neural network trained on numerous examples from real life. HTTP API is provided to exchange data with the service.

Yandex SpeechKit

Smart_Engines

Computer vision is a system based on OCR, computer vision and machine Learning technologies. API solutions will allow to use the system for the following tasks: fuzzy search of similar images, motion detection, non-format document scanning on mobile devices, holographic elements detection, determining the vehicle type, traffic lights color recognition and tags recognition. 

 Information is recognized in real time and the system does not require connection to the internet. On the basis of the technology <a href="https://itunes.apple.com/ru/app/smarthelper/id972688836?ls=1&amp;mt=8" target=«_blank»> a mobile app</a> SmartHelper (for IKEA) for iOS and Android was created, it can recognize the set number, row and place of the product in the warehouse from IKEA tags.

System based on OCR technologies, computer vision and machine learning

machine_learning

Computer Vision

Smart PlateReader ensures license plate recognition on mobile devices. The technology allows to recognize car plate number in smartphone and tablet apps very quickly. Only the operative memory of the device is used during recognition.

The time for recognition is estimated at 50 ms - 2 s (depending on the device). The area for implementation of the solutions based on Smart PlateReader are insurance companies, social and state services that detect traffic violations, car sharing services and others.

Smart PlateReader

Smart ArchiveHelper is a range of technological solutions which allows to create an online archive with automatic document classification and checking the completeness of the set. It also allows to process the existing image archives, reducing the size (to up to 90%) and increasing the informative value of the contained elements.

System of document preparation for further storage

Smart ArchiveHelper

Smart CardReader allows to automatically recognize the number, date and name on all types of bank cards. The system does not require connection to the internet and does not store or transmit the recognized bank card data. Solutions based on Smart CardReader can be used to recognize club, discount, transport and other types of plastic cards.

Integration of Smart CardReader into payment services make it possible for the users to automatically enter bank card data for payments, money remittances and other financial operations in mobile apps and on websites. Processing time is around 1 second, and the number recognition accuracy is not less than 99.61%.

<a href="https://apps.apple.com/us/app/smart-code-engine/id1597490872">Mobile app</a>

Smart CardReader

The system identifies clients by checking their identity documents in mobile, desktop and server apps. The solution extracts data in the video stream from the passport of the Russian Federation, SNILS, driving licenсe, vehicle registration certificate; the system also recognizes photos and document scans. &nbsp;

Smart IDReader determines the type of the document, recognizes text and machine readable zones (MRZ), finds a photo and a signature. Moreover, Smart IDReader allows to recognize 1D barcodes.

The system monitors the process of recognition and warns the user about the possible mistakes. Smart IDReader uses operative memory (does not create copies or transmit data for processing via internet to the cloud or server)

In June 2018 <a href="«http://smartengines.ru/news74/">a new version</a>&nbsp;of the system, Smart IDReader 2.0 was released .&nbsp; It was created using deep sparse discrete neural networks, it can recognize official documents, A4 format documents in a video stream or image and deals with random camera angles. The new version provides an opportunity to recognize a handwritten passport of a citizen of the Russian Federation.

<a href="https://play.google.com/store/apps/details?id=biz.smartengines.smartid" target="«_blank&quot;">Mobile app</a>

System for identity documents recognition

identification

Smart IDReader

VisionLabs

Face DJ is based on 3D face model reconstruction technology and can transfer it to virtual reality. 3D model reconstruction from a single Image is available due to the technology that defines key points on the face and transfers them to the adaptable virtual head model.

Individual face features are defined in the image, and the virtual model adapts to them. Face key points definition algorithm includes 2 stages. Firstly, the face is detected in the image, then it is extracted from the image and complemented with a set of pre-defined key points. Machine learning is used in both stages. 

<a href="https://itunes.apple.com/ru/app/face-dj/id1208394008?mt=8" target=«_blank">Mobile app</a>

App for 3D model creation and transferring images to virtual reality

Face.DJ

Luna Platform is based on LUNA SDK - Engine for face recognition. The system includes several modules: Broker web server, which ensures interface with external systems via HTTP protocol; Extracto handler, which extracts face descriptors; Matcher handler, which receives requests for descriptors matching. An additional Queue Manager ensures data exchange between Broker module and Extractor and Matcher handlers. DBMS Aerospike NoSQL module is used to create a storage for descriptors and their lists. The modules work independently from each other. The system uses RESTful Web API to transmit commands and exchange data.

 The main Application level protocol is HTTP 1.1, data exchange is initiated by an HTTP request. The implementation of neural networks allows to process a face in a fraction of a second, and the optimized sublinear search makes it possible to identify a person in the databases. The descriptors data is de-identified

System of data management and person verification and indentification

Tech Solutions

Banknote identifier

VOCORD MicroCyclops

ABBYY FlexiCapture

add your project!
fill out the form

Yandex SpeechKit

Computer Vision

Smart PlateReader

Smart ArchiveHelper

Smart CardReader

Smart IDReader

Face.DJ

LUNA Platform

Tech Solutions

Banknote identifier

VOCORD MicroCyclops

ABBYY FlexiCapture

add your project!fill out the form

Yandex SpeechKit

Computer Vision

Smart PlateReader

Smart ArchiveHelper

Smart CardReader

Smart IDReader

Face.DJ

LUNA Platform

add your project!
fill out the form