Computer Vision

The development of machine learning, computer equipment, digital video and photographic technologies, and the Internet have, in recent years, led to a significant improvement in computer vision systems' performance quality. For some tasks, such as face recognition, work is already being carried out at quite a high level.

As a result, vision algorithms have been implemented in numerous models of digital cameras, there are a variety of programs for managing collections of photographs and so on. In recent years, and even months, significant breakthroughs have been achieved in multiclass categorization, with the number of classes available in current systems running into the thousands, and the difference between several different classes can be reduced to very fine details.

However, despite all this success, many problems in computer vision, linked in part to the large archive of images, deep understanding of images, robotics, video surveillance and so on are still a long way from having satisfactory solutions.

A wide variety of computer vision technologies are currently in use at Yandex, with the most striking examples being in the fast-search through images and the filtering of pornographic content. The seminars will host presentations covering developments participants, and invited speakers, are currently working on, and there will also be discussions of recently published papers from leading conferences.

See Seminar webpage for more detailed information.

Seminar moderators: Artem Babenko, Anton Konushin, Viktor Lempitsky, Anton Slesarev.