Deep Learning for stereo matching and related tasks (Talk)
Recently, deep learning proved to be successful also on low level vision tasks such as stereo matching. Another recent trend in this latter field is represented by confidence measures, with increasing effectiveness when coupled with random forest classifiers or CNNs. Despite their excellent accuracy in outliers detection, few other applications rely on them. In the first part of the talk, we'll take a look at the latest proposal in terms of confidence measures for stereo matching, as well as at some novel methodologies exploiting these very accurate cues. In the second part, we'll talk about GC-net, a deep network currently representing the state-of-the-art on the KITTI datasets, and its extension to motion stereo processing.
Biography: I studied Computer Engineering at University of Bologna, where I received my bachelor degree in October 2012 (work of thesis: software deployment and porting for a custom FPGA stereo camera designed at University of Bologna) and my master degree in October 2014 (work of thesis: development of a fast and accurate stereo algorithm based on tree filtering and plane fitting), both with honors. During my master thesis in 2014 I had an internship of 6 month at Aquifi, Palo Alto, where I worked under the supervision of Professor Roberto Manduchi on real-time stereo algorithms. Since November 2014, I'm a PhD student at University of Bologna, under the supervision of Professor Stefano Mattoccia, and a teaching assistant for the bachelor degree course of Computer Architectures. My main research interests are stereo matching and deep learning, I'm also interested into embedded and reconfigurable systems for stereo.
