Applied Projects and Deep Fakes: How Computer Vision Is Taught at HSE University

Applications for the HSE University Master of Computer Vision, the only English-language online computer vision programme in Russia, are open until August 10. The programme has been developed by researchers of the Faculty of Informatics, Mathematics, and Computer Science at HSE University in Nizhny Novgorod together with researchers in the field of computer vision from leading companies in the industry: Huawei, Itseez3D, Intel, Harman, Xperience.ai , Sber, Newstream and Deelvin Solutions. Andrey Savchenko, Academic Supervisor of the programme and Professor at the Department of Information Systems and Technologies, told the HSE News Service about how teaching competencies in the field of computer vision changes our view of the world.

Andrey Savchenko, Academic Supervisor of the programme 'Master of Computer Vision'

What is the main goal of the programme?

The design of the programme considers the interests of applicants with mathematical, technical or IT backgrounds who want to immerse themselves in the latest research in the field of artificial intelligence, data analysis and machine learning (including deep learning). Students will get acquainted with modern practices in the field of computer vision, including methods of processing, analysis and synthesis of images and videos.

In addition to a strong academic foundation, the programme focuses on current industry cases

Students not only receive theoretical knowledge from HSE University professors, but also solve tasks presented by employees of the programme's industrial partners.

What disciplines does the programme offer?

In order to acquire the necessary knowledge and skills, students will take courses in mathematics for computer vision, 2D/3D image processing, deep learning in the field of computer vision, software development for CV projects, and more.

Each course involves applied projects. The practical part is based on the business experience of IT companies, and all the tasks are study cases provided by our teachers and based on their industry tasks. These include, for example, facial recognition for video surveillance systems, client-server applications for object detection and semantic segmentation, photo analysis and modification applications on mobile devices, image generation (including deep fakes), diagnosis of diseases by images in computed tomography and much more. As a result, our graduates enter the labour market with an impressive portfolio of completed projects. Thanks to close contact with leading CV specialists, our graduates are well-adapted to the professional environment.

What career opportunities do graduates have?

The online master's programme trains highly qualified specialists who can work in any projects related to computer vision (including the development and implementation of multimodal systems) and have skills and tools for solving a wide range of tasks. For example, they can get involved in object recognition, the creation of 3D reconstructions and photo filters, mobile applications for recognising objects in photos and videos, the introduction of ML in production, retail, medicine, banking, agriculture, etc. Graduates can work positions such as Computer Vision Software Engineer, Perception Engineer, 3D Perception/Computer Vision Algorithm Engineer, Computer Vision Testing Engineer, Computer Vision Specialist, Data Processing Specialist, Machine Learning Engineer.

Is it convenient to study such a complex programme online?

The online format of the programme allows students to combine work with study and get a degree from one of the leading universities in the world without moving to Nizhny Novgorod. The study process is implemented on the Smart LMS platform and takes 20–30 hours a week.

Online students are always in touch with the study office, and quick feedback is available in chat rooms, forums, and live sessions

In addition, the programme curriculum provides weekly live consultations—dedicated times to speak with teachers outside lectures.

The programme combines synchronous and asynchronous class formats, and students study two courses simultaneously. Basic adaptation courses in mathematics and information technology are combined with a specialised block covering a modern theoretical and instrumental background in the field of computer vision. It is focused on the study of theoretical approaches and means of solving applied problems that arise during work. Interactive work involves live webinars—three meetings during each course. Each module includes six webinars between groups and teachers.

English-language instruction contributes to the creation of an international student community

Students from Saudi Arabia, Vietnam, China, Romania, France, Kazakhstan and, of course, Russia study on the programme.

On July 27, you are holding the Sber AI Lab Medical Imaging Group ongoing projects webinar. What will it be dedicated to?

Teachers of the Modern Tasks of Computer Vision course from the Sber Artificial Intelligence Laboratory will speak about a number of applied tasks in medical image recognition. In particular, they will give an overview of classical approaches to image processing for computed tomography and electrocardiograms.

We plan to thoroughly discuss modern neural network methods. In theory, you can simply collect a large number of medical images and carefully mark them up—for example, select a tumour area then train a suitable neural network. But industry representatives know that in practice, this approach does not work; it is extremely difficult and expensive to perform accurate markup of images. Therefore, the webinar will consider self-learning models that do not require a large amount of marked-up data.