Thomas Brox

Title: Will Deep Learning Solve Computer Vision?

Keywords: Convolutional networks, segmentation, optical flow

Abstract: Deep learning based on convolutional network architectures has revolutionized the field of visual recognition in the last two years. There is hardly a classification task left, where ConvNets do not define the state-of-the-art. Outside recognition, deep learning seems to be of lesser importance, yet this could be a fallacy. In this talk I will present our recent work on convolutional networks and show that they can learn to solve computer vision problems that are not typically assigned to the field of recognition. I will present a network that has learned to be good on descriptor matching, another one can generate images from abstract features, one network is good on image segmentation, and there is even a network that has learned to estimate optical flow. The recent success provokes the question whether deep learning is finally the technique that will allow us to solve computer vision. Also, can it help us understand how our biological vision system works?

Bio: Thomas Brox received his Ph.D. in computer science from the Saarland University, Germany in 2005. Afterwards he joined the Computer Vision Group at the University of Bonn as a postdoctoral researcher. He headed the Intelligent Systems Group at the University of Dresden as a temporary faculty member for one year. After two years as a postdoctoral fellow in the Computer Vision Group of Jitendra Malik at U.C. Berkeley he moved to the University of Freiburg, where he is heading the Computer Vision Group. Prof. Brox is associate editor of the IEEE Transactions on Pattern Analysis and Machine Intelligence and the International Journal of Computer Vision. He has been an area chair for ICCV 2011, ACCV 2014, ECCV 2014 and ICCV 2015 and reviews for several funding organizations. He received the Longuet-Higgins Best Paper Award in 2004 and the Koenderink Prize for Fundamental Contributions in Computer Vision in 2014 for his work on optical flow estimation. In 2011 he was awarded an ERC starting grant.