AI & Fundamentals
Learning to Anticipate - Alex Schwing, Assistant Professor, University of Illinois
DATE: Mon, December 16, 2019 - 3:00 pm
LOCATION: ICCS - X836, ICICS Computer Science, 2366 Main Mall, Vancouver, BC
Despite significant progress in recent years, autonomous agents like speakers or cars are far from participating robustly and safely in our environment, largely because they lack an ability to anticipate. We argue that this is due to four reasons: (1) holistic object reasoning is still at its infancy, e.g., because of occlusions; (2) inferring of interactions between observed objects is hard because little data is available; (3) capturing of revealing real-world priors is challenging because of a high-dimensional setting; and (4) understanding ambiguity is tough due to an exponential number of possibilities.
In this talk we present vignettes of our research to address those challenges. Specifically, we first discuss a new method for weakly-supervised instance-level video object segmentation: given the objects of interest in the first frame of a video we describe our matching based approach for tracking and detecting objects across frames. We then illustrate challenges due to occlusions and present our recent work towards amodal segmentation. In the second part we present our recent work on developing systems that understand ambiguity in question answering and captioning.
Alex Schwing is an Assistant Professor at the University of Illinois at Urbana-Champaign working with talented students on computer vision and machine learning topics. He received his B.S. and diploma in Electrical Engineering and Information Technology from Technical University of Munich in 2006 and 2008 respectively, and obtained a PhD in Computer Science from ETH Zurich in 2014. Afterwards he joined University of Toronto as a postdoctoral fellow until 2016. His research interests are in the area of computer vision and machine learning, where he has co-authored numerous papers on topics in scene understanding, inference and learning algorithms, deep learning, image and language processing and generative modeling. His PhD thesis was awarded an ETH medal. For additional info, please browse to http://alexander-schwing.de.
Associate Professor Leonid Sigal, Computer Science, UBC