Mail.ru Vision

Mail.ru Vision: API for integration of face recognition and objects based on machine learning

Automation of any human work which is based on visual reconciliation of the original (or the preset conditions) with a photo, video or images in real time.

2020: Testing of a sensing technology of queues based on Vision

On February 20, 2020 Mail.ru Group announced that with the assistance of specialists of MRG Tech Lab (the direction of the company which creation it was announced on March 1, 2019) tests a sensing technology of queues. Computer vision of Vision in real time analyzes how many people are in different sections of a frame. If queue increases, the program reports about it on the person on duty panel.

The basis of Vision is formed by technologies of self-training neuronets. The artificial intelligence analyzes records of the systems of photo and video surveillance and will recognize on them silhouettes of people. Vision less than in a second defines how many people are in a frame.

The solution is developed for platforms where a large number of people concentrates: airports, shopping centers, restaurants, skating rinks, alpine skiing complexes, festivals and city holidays. Vision successfully copes even in situations when silhouettes on video block each other.

Computer vision not only fixes the number of people, but also monitors their distribution by different parts of a frame. For example, if many visitors gathered in one of departments of supermarket, Vision will notify the administrator that there it is necessary to bring more goods or to invite still consultants.

According to developers, Vision equally effectively works both at the open area, and in premises. A system constantly studies and saves the high accuracy of recognition even when shooting in difficult conditions: during a rain, snow, at bad lighting. For example, she is capable to analyze queue to the elevator in ski resort.

Specialists of MRG Tech Lab can adapt technology for business objectives of the customer. Pilot projects are already started in the transport and industrial companies.

2018: Start of technology of computer vision of Vision

On February 20, 2018 Mail.Ru Group started on the market of B2B-services of technology of computer vision of own development - Vision. The technology underwent testing on the products Mail.Ru Group: on mobile application of Cloud of Mail.Ru for face recognition in photos of the user.

Accuracy of recognition reaches nearly 100% (a photo - iqubesdp.com)

The Vision technology is offered to potential consumers as PaaS-service within the Mail.Ru for business platform, follows from the message of the company.

Vision is founded on machine learning technology. Using neural networks it provides high quality and high speed of recognition of different objects on photo and video files both in offline, and in online the modes, affirms as the press release of Mail.Ru.

Vision can be used for automation and gain of the check mode at the enterprises, in retail a system will allow to raspoznatvat regular customers and unwanted guests, to control the calculation of goods, on production sites with its help it will be possible to automate a number of control functions.

The market of computer vision is at an early stage of development, are sure of developer company. According to Mail.Ru, in 2017 its volume was about 700 million rubles. Experts predict its growth to 2 billion rubles within two-three years.

{{the quote|Supply and demand only begin to form. The most popular product in this market – recognition of visitors in shops and banks, in many respects because it is the most obvious method of application of machine vision. We are going to develop the market actively: specialists of Vision advise partners, finding cases in which the technology of computer vision from Mail.Ru Group will help to increase efficiency and to reduce business costs, – the press service of the company of fishing of the head cloud and Mail.Ru Group business services of Egor of Ganin.} brings }

On the website Vision it is reported that the sensing technology of persons and objects based on machine learning can be adapted for individual needs of business. The main objective which a system will help to solve, - reduction of a human factor in business process: Vision automates any human work which is based on visual check of the original with what occurs in the photo, video or in production process. It allows to reduce the probability of an error and to reduce choice reaction time.

Via the program interfaces API facial recognition, determination of scenes and objects, determination of quality the image, tracking of people, recognition of celebrities, text recognition on the image, moderation of images, determination of numbers of cars, identification of defects on production, search of sights, increase in permission and detection of cars is already available.

According to the company, the average speed of the analysis of the image is 100 milliseconds, and recognition accuracy – 98%.