Interview with Prof. Daniel Cremers

hello this is Stefan morta I'm reporting from the International Conference in consumer electronics in Seoul Korea this is the inaugural conference and our inaugural keynote speaker professor Amos from the Technical University of Munich is here with me and he gave us a very interesting presentation on his work professor Kramer's my first question to you is you back to your background is in mathematics and physics and now you're working in image processing so you know it's sort of a unusual combination can you tell us how you actually sort of moved into this area yes so the reason why I moved into image processing and computer vision is because I loved mathematics but I wanted to do things that really impact our society that impact everyday life and this is certainly something that we see now the impact we have with the technologies we develop is quite substantial and this is the one thing that I find quite fascinating in this work yes it is a fascinating field that you're working on particularly from your presentation I found the autonomous driving applications very interesting would you mind to elaborate about what you're doing and how you see the future in that area so when we develop some of the core technologies to enable the driver assistance and self-driving cars so it's all about getting from cameras and other sensors to an understanding of what happens around the car what happens in particular in front of the car in the driving corridor you want to see obstacles you want to avoid obstacles you want to save lives and you want to on the long run be able to make predictions about what's gonna happen in the next few seconds in front of my car so that the car can react autonomously and in that way save lives it's a very challenging problem but we're making quite some progress there yes that's a indeed a very challenging problem and noticed in your presentation you talked also about how you combine different technologies to identify problem areas like totally a white background background and we've heard famously that the tesla car crashed into a truck because it was a blue background something like that it could identify that it was an object yes so it is a challenging problem and one of the things we noticed in the tense Tesla incident that it is a is a challenge that really involves a substantial possible damage I believe though we are on a very good track and one of the great things about this technology is the technology will become safer and safer when I'm the machines learn they improve with every new generation they improve and with every with every problem solved all the cars can profit you know if one driver makes a mistake a human driver the other drivers don't improve if one self-driving cars makes a mistake and learns from me all existing self-driving cars will profit from that and so there will be an overall systemic improvement in the technology to a degree that on the long-run self-driving cars will make driving safer than human drivers yeah that's very interesting in peripheral research area but also during your presentation you talked about different areas that you'll be working on one of the areas you elaborate around was neural networks and you also mentioned that you saw step back for a while from artificial intelligence methods and now you're rekindling it can you tell us a bit about that so the neural networks have become quite popular again in recent years in the computer vision area because they allow to solve certain computer vision problems at a level that is unprecedented in terms of performance recognizing objects in images recognizing objects in videos segmenting images into semantic components on all of these challenges neural networks seem to be outperforming previous techniques and algorithms so there they do hold quite some promise in in in terms of understanding the world from images yes that's true last question I'd like to ask you is the medical field is also advancing very fast and a lot of consumer techniques and vision seeing areas of finding very useful applications in the medical field do you see a big future in that oh yes definitely medical image analysis is one of the biggest areas of image processing and computer vision in terms of the market value definitely but also in terms of the importance for society the the ability to diagnose certain illnesses and diseases from imagery data is very important we're scanning people these days with more and more imaging technologies two-dimensional three-dimensional scans scans over time we can get 4d ultrasound ultrasound volumetric ultrasound over time there's so much data that we can acquire one of the core challenges that it remains is to interpret that data and make sense of it automatically to date we still rely heavily on medical experts that manually look through images to see whether there is a tumor or something else we're hoping that on the long run a lot of this manual human labor can be replaced by machines that will then ideally perform better and more robustly more with less errors than the human observer will yes so it is very challenging opportunities ahead for us but put to improving the area of vision processing I'd like to thank you for your very interesting keynote that you presented today and for the viewers of this clip here I believe the keynote will be available on CCC stock YouTube TV so I recommend highly you come back and have a look at that video clip Thank You professor Kramer's thank you so much

Channel: IEEE CTSocTV

Views: 1,633

Rating: 5 out of 5

Keywords: ICCE, IEEE, 2016, Asia, Daniel Cremers, Stefan Mozar, Interview

Id: u9jz124TQxs

Length: 6min 31sec (391 seconds)

Published: Thu Nov 24 2016