Yandex Vision

2020: Integration into the ElectroNeek RPA platform

On July 7, 2020 the ElectroNeek Robotics company announced release of release of ElectroNeek 3.0 Siberia. Within release of the company can not only automate routine working tasks, but also automatically find inefficient business processes both in corporate network, and on the Internet which optimization will allow to increase productivity of work and to reduce operating costs. By means of technology of the analysis of images Yandex Vision users also have an opportunity to process data of private papers, to work with scans and other formats of data that expands possibilities of application of ElectroNeek in work of the companies. Read more here.

2019: Start of Yandex Vision service

On April 22, 2019 it became known that "Yandex" started on the cloud platform "Yandex.cloud" Yandex Vision service. This service is intended only for software developers and offers a set of technologies of computer vision of "Yandex" for recognition of information on images. It integrates technologies of optical character recognition (OCR, Optical character recognition), pattern discriminations and detection of persons.

Earlier Yandex Vision was used only in internal projects, in particular in the search engine and the translator of "Yandex".

Representatives of "Yandex" among competitors of Yandex Vision called solutions Google and Abbyy. Really, Google Cloud Vision API Abbyy OCR Cloud SDK are also similar to destination and functionality to a prelstavlenny cloud service of "Yandex".

Yandex Vision, according to representatives of the company, is easily integrated with applications via the REST API or gRPC API interfaces, and implementation of integration from the developer does not require experience in the field of machine learning.

Using Yandex Vision service developers can add to own products function of text recognition in documents and images. If the digitized image (the photo, the scanned page) contains the text, then Yandex Vision will recognize it. For April, 2019 service supports more than 40 languages, including languages of the CIS countries.

Such indicators are reached at recognition of typewritten sources of high quality. According to documentation published on the website "Yandex.cloud", service works with some restrictions, for example, the algorithm Yandex Vision not in the best way will recognize hand-written, vertical or very large text.

The main competitive advantages of service in "Yandex" consider low price – about 0.1 rub for the page of the recognized text, high speed of work of an algorithm (1–2 seconds on the page of the document) and also high quality of a final result. According to "Yandex", for a set of the photographed and scanned documents the Russian and English texts are recognized with a medium accuracy about 96%, Kazakh – 94%.

In addition to Yandex Vision text recognition also allows to define presence of persons on the image and their quantity and also to receive their coordinates for postprocessing. For example, it is possible to find all persons in the photos loaded by the user and to suggest to mark out familiar people.

The technology of pattern discrimination will help to analyze content which is loaded by users, for example, to reveal copyright violation on existence of watermarks or to filter an inappropriate content.

Possibilities of Yandex Vision on detection of persons and pattern discrimination for April, 2019 are at a preview stage therefore are provided free of charge.

Representatives of "Yandex" told that they are going to work on increase in number of services of computer vision both for document handling, and for work with photos and a video flow^[1].

Notes