Machines Can See: International Summit on Computer Vision
On June 9, a summit was held on computer vision and deep learning ‘Machines can see’, organized by Sistema VC, Visionlabs, and the Strelka Institute. Dmitry Vetrov and Anton Konushin, staff members of HSE Faculty of Computer Science, were among the organizers of and speakers at the conference.
Computer vision is a branch of science relating to artificial intelligence, the theory and technology of developing systems that can extract meaningful information from images. The development of this area has a huge impact on robotics, developing unmanned vehicles, augmented reality, medicine, and many other industries.
This is the first time that the international summit on computer vision ‘Machines can see’ has taken place in Russia and the CIS countries. Participants’ reports focused on the latest trends in the field of artificial intelligence, deep learning algorithms, neural networks, the problem of object recognition and dynamic scenes, as well as their practical application.
World-leading scholars took part in the event: Jean Ponce, MIT and Stanford Professor, author of the book 'Computer vision: a modern approach' ; Soumith Chintala, Artificial Intelligence Research Engineer at Facebook; Cordelia Schmid, INRIA Research Director (Grenoble, France), who was awarded Humboldt Research Award for her contribution to the study of computer vision; Alexei Efros, Professor at EECS Department at UC Berkeley; and Jiri Matas, Head of the Center for Machine Perception at the Czech Technical University in Prague, who has written over 300 works on image tracking and recognition.
Russian developments in the field of computer vision and machine learning were presented by Dmitry Vetrov, Senior Researcher at Yandex and Head of the International Laboratory of Deep Learning and Bayesian Methods at the Faculty of Computer Science, Konstantin Lakhman, Head of Deep Learning Team (Computer Vision) at Yandex; Victor Lempitsky, Associate Professor at the Skolkovo Institute of Science and Technology and Head of the Computer Vision Group; and Alexander Chigorin, Head of VisionLabs research projects.
Michael Figurnov and Arsenii Ashukha, staff members of the International Laboratory of Deep Learning and Bayesian Methods, and Anna Sokolova, doctoral student in Computer Science, also presented their work in the poster session.
Dmitry Vetrov, Head of the International Laboratory of Deep Learning and Bayesian Methods
‘The event was interesting and well received by the audience. Over 1,000 people took part in the workshop, and they were not all specialists in machine learning, indicating that the topic is becoming more interesting to a general audience.
I found it interesting to hear reports by my colleagues; many of whom are world-leading researchers. For example, Soumith Chintala is chief developer of the library PyTorch, launched in February 2017, and all leading research groups in the world engaged in deep learning are switching to it.
My report was dedicated to the successful application of modern Bayesian methods to the problems of learning and regulating deep neural networks. The key mathematical tool was developed in 2015-2017. It eliminates the redundancy of modern neural network architectures, rejecting up to 99.9% of the weights of the network with no loss in quality and general capacity. The report was based on an article prepared by international laboratory staff and was recently accepted for the major international conference on machine learning ICML2017, which will be held in August in Sydney.
For more information on the event, presentations, speakers, and videos, please visit the conference website.