When humans see a scene they can quite effortlessly visually separate individual objects present in the scene. For a computer, the ultimate but very challenging goal is to achieve the same without the help of humans.
When humans see a scene they can quite effortlessly visually separate individual objects present in the scene. For a computer, the ultimate but very challenging goal is to achieve the same without the help of humans. I will consider a slightly simpler task, where the user is allowed to give hints about what should be segmented. This task is called “Interactive Image Segmentation”. I will look at this problem from two different angles. Firstly, I will describe an approach to this problem which involves a joint optimization over appearance and shape.
I will present our results from a sequence of papers in which we over time improved both the model and the optimization method. We shall focus on methods aiming to find a globally optimal solution. Secondly, I will examine the role of the user. The ultimate goal is to design a system with which the user obtains an acceptable result in as little time as possible. We show that this is possible by integrating the user into the computational model and that it results in an improved system. In fact, this method, called GrabCut, is the basis of the foreground extraction tool that ships with the Microsoft office products.
Carsten Rother received a PhD degree in 2003 from the Royal Institute of Technology Stockholm, Sweden. From 2003 until 2013, he has been a researcher at Microsoft Research Cambridge, United Kingdom. Since October 2013 he is full professor at TU Dresden, heading the Computer Vision Lab Dresden. His research interests include scene recovery and understanding”, and optimization and learning in probabilistic graphical models. He has published more than 120 articles (H-index 52 according to Google Scholar) in international conferences and journals. He won awards at a number of conferences (ACCV, CVPR, BMVC, CHI, CVPR) and the DAGM Olympus prize. He has influenced various Microsoft products, in particular GrabCut for Office 2010 and AutoCollage. He serves in programme committees of major conferences (e.g., SIGGRAPH, ICCV, ECCV, CVPR, and NIPS). He received the ERC Consolidator Grant in 2014.
Its program consists of a one-hour lecture followed by a discussion. The lecture is based on an (internationally) exceptional or remarkable achievement of the lecturer, presented in a way which is comprehensible and interesting to a broad computer science community. The lectures is in English.
The seminar is organized by the organizational committee consisting of Roman Barták (Charles University, Faculty of Mathematics and Physics), Michal Chytil (Czech Academy of Sciences, Computer Science Institute), Pavel Kordík (Czech Tech. Univ., Faculty of Information Technologies), Jan Kybic (Czech Tech. Univ., Faculty of Electrical Engineering), Michal Pěchouček (Czech Tech. Univ., Faculty of Electrical Engineering), Jiří Sgall (Charles University, Faculty of Mathematics and Physics), Vojtěch Svátek (University of Economics, Faculty of Informatics and Statistics), Michal Šorel (Czech Academy of Sciences, Institute of Information Theory and Automation), Tomáš Werner (Czech Tech. Univ., Faculty of Electrical Engineering), and Filip Železný (Czech Tech. Univ., Faculty of Electrical Engineering)
The idea to organize this seminar emerged in discussions of the representatives of several research institutes on how to avoid the undesired fragmentation of the Czech computer science community.