Speakers

Dr. Liu is a professor of EECS at Penn State University (PSU), trained in physics/EE (B.S., Beijing, China), computer science (M.S., MA, USA), and theoretical robotics/AI (Ph.D. MA, USA; Postdoc, France). With an NSF (USA) research-education fellowship award, she spent one year at DIMACS (NSF center for DIscrete MAthematics and Theoretical Computer Science) before joining the faculty of Robotics Institute, Carnegie Mellon University for ten years. Currently at PSU, she is the director of the Human Motion Capture Lab for Smart Health and co-directs the Lab for Perception, Action and Cognition (LPAC). A central theme of Dr. Liu's research remains “computational regularity” (funded by USA NSF, including an INSPIRE grant) with diverse applications in robotics, human/machine perception, and human health. Her industrial visits resulted in two granted patents on applying computational symmetry to urban scenes. Dr. Liu led three international competitions on Computer Vision algorithms for Detecting Symmetry in the Wild (ICCV 2017, CVPR 2013, CVPR 2011). She served as a co-program chair for CVPR 2017 and WACV 2019, area chairs for multiple CVPR/ECCV/ICCV/MICCAI conferences, and an associate editor for IEEE Transaction of Pattern Analysis and Machine Intelligence (PAMI).

Title: Localization with Contacts: Physical or Virtual

Abstract: Data-driven learning-based methods have demonstrated various levels of success in robotics and computer vision/graphics applications, yet a link remains missing: are there any principles to be based on or to be learned in modern day sensor-based robotics to connect the data and fundamental mathematics? In this talk, I shall explore this topic from a historical perspective as well as from state-of-the-art localization issues faced by robotic manipulators, humanoid robots and humans alike.   

Dr. Crandall is a full Professor in the School of Informatics, Computing, and Engineering at Indiana University, where he directs the IU Computer Vision Lab. He is on the faculty of the Computer Science, Informatics, Cognitive Science, and Data Science programs, and he is adjunct in the Department of Statistics. His work has been funded by grants and contracts from the National Science Foundation (including through a CAREER award), Facebook, Google, Kodak, the Department of Defense, and Indiana University.  Before joining IU, he was a graduate student and postdoc in Computer Science at Cornell University. Before Cornell, he spent two years in the research labs of Eastman Kodak Company.

Dr. Daniel Aliaga received the B.S. degree from Brown University and the M.S. and Ph.D. degrees from The University of North Carolina at Chapel Hill (UNC Chapel Hill). He is Associate Editor for IEEE TVCG and for Visual Computing Journal (previously for Computer Graphics Forum and Graphical Models) and PC member for SIGGRAPH, CVPR, ICCV, Eurographics, AAAI, I3D, IEEE Vis. He has received a Fulbright Scholar Award, a Discovery Park Faculty Research Fellowship, and his PhD advisees have received a total of 11 Purdue fellowships/grants. He is a member of ACM SIGGRAPH and ACM SIGGRAPH Pioneers, and has multiple times been Chair of Faculty Diversity for College of Science at Purdue.  Dr. Aliaga has obtained 28 external peer-reviewed grants totaling $40M (and is PI on 19 of them). Funding sources include NSF, IARPA, USDA, Internet2, MTC, Google, Microsoft, and Adobe. Dr. Aliaga is a pioneer in the area of inverse procedural modeling for urban spaces, architecture, and many other areas. 

Title: Deep Visual Computing for Urban Modeling and Design

Abstract: Dr. Aliaga converts incomplete unstructured geometric and semantic data of 3D urban environments into organized and easily editable representations. These representations can help improve the complex urban ecosystem and provide "what-if" exploration of sustainable urban designs, including integrating urban 3D modeling, simulation, meteorology, vegetation, and traffic modeling. This talk will provide an overview of the methodologies he has developed including deep learning approaches and forward and inverse procedural modeling algorithms to create/modify 3D and 2D urban models. His work has, and is, solving problems such as: “given completely disorganized geometric data, can an underlying organized structure be inferred?”, “given a fragmented and highly occluded observation of an urban structure, can the full structure be inferred?”, “given extremely low-resolution blurry observations, can the underlying ground truth far beyond what is visible be inferred?”, and “given only a few-percent of a structure, how can you provide a 100% version of the structure?” Aside from publications, the work has also resulted in deployed cyberinfrastructure prototypes using ground-level imagery, aerial imagery, satellite imagery, GIS data, and, PhD committee memberships in computer science, atmospheric sciences, urban planning, and forestry, numerous international talks, TEDx talk, and driven-forward global grassroots efforts (e.g., WUDAPT).

Dr. Xinyu Du received the B.Sc. and M.S. degrees in Automation from Tsinghua University, Beijing, China, in 2001 and 2004, respectively, and the Ph.D. degree in electrical engineering from Wayne State University, MI, USA, in 2012. He has been working in General Motors Global R&D Center, Warren, MI since 2010, and currently holds the staff researcher position in vehicle systems research lab. His work on integrated starting system prognosis has been implemented in more than 2 million GM vehicles and received the best innovation award (Boss Kettering Award) from General Motors in 2015. This technology is one of the supporting practices that won the INFORMS award for General Motors in 2016. Two of his papers in vehicle health management area were recognized as the best conference paper award in 2019 and 2020. His research interests include fuzzy hybrid system, vehicle health management, machine learning, autonomous vehicles, and battery management system. He has been serving as an associate editor for Journal of Intelligent and Fuzzy Systems from 2012.

Dr. Yao Hu is a Staff Researcher at Propulsion Systems Research Lab, Global Research and Development, General Motors. He has been working as a researcher at R&D GM since 2016. He received a Ph.D. in electrical engineering from University of Kentucky in 2013, an M.Eng. from Shanghai Institute of Technical Physics, Chinese Academy of Sciences in 2007, and a B.Sc. from Fudan University, Shanghai, China, in 2004. His research interests include sensor alignment, localization and perception in autonomous driving, vehicle fault diagnostics and prognostics, and vehicle telematics applications. Currently, he leads the algorithm development of online LiDAR-to-vehicle alignment and works on online camera-to-vehicle alignment for Ultra Cruise in GM vehicles.

Title: Online LiDAR Alignment (combined talk with Dr. Xinyu Du)

Abstract: GM has been launching autonomous driving features such as Super Cruise and Ultra Cruise. The LiDAR sensor, as a part of the Ultra Cruise system, requires accurate and robust extrinsic calibrations. Over the past few years, we have been evaluating and developing various methods for online LiDAR alignment using technologies in the visual odometry area while considering the limitation of the autonomous driving application. These methods include aggregation-based, hand-eye, and prior-knowledge-based methods. They show advantages and limitations under different scenarios. These studies provide a solid foundation to develop the production solution for GM Ultra Cruise.  

Dr. Enrique Dunn is an associate professor in the Department of Computer Science. His research focus is on 3D Computer Vision, investigating the geometric and semantic relationships among a 3D scene and a depicting set of images. Dr. Dunn earned a degree in computer engineering from the Autonomous University of Baja California (Mexico) in 1999. He completed a master’s degree in Computer Science in 2001 and a doctorate in Electronics and Telecommunications in 2006, both from the Ensenada Center for Scientific Research and Higher Education (Mexico). During his doctorate studies, Dr. Dunn carried out research while visiting the French Institute for Research in Computer Science and Control in Rocquencourt. He joined the Department of Computer Science of the University of North Carolina at Chapel Hill as a visiting scholar in 2008, after being awarded a one year Postdoctoral Fellowship for Studies Abroad by the National Council for Science and Technology (Mexico). He remained with UNC-CH CS Department as a postdoctoral researcher until he became a research assistant professor in 2012. Dr. Dunn has authored over 40 papers in international conferences and journals. He is a member of the Editorial Board of Elsevier Journal of Image and Vision Computing. ​​

Title: DATA-DRIVEN GEOMETRIC WORKFLOWS FOR VISUAL LOCALIZATION

Abstract: Visual localization aims to estimate the spatial relationship between a camera and its environment based on visual media captures. However, the increasing need for robustness in challenging environments and the requirement to solve semantic-centric problems has presented challenges for geometric localization workflows. Conventional methods rely on hand-crafted heuristics that struggle to meet the growing demands of such environments. On the other hand, emerging deep learning methods face issues with generalization and interpretability due to their geometry-agnostic nature. To address these challenges, we explore the use of hybrid localization workflows that leverage both geometric and data-driven priors.

Our recent research highlights how such alternative formulations advance the  robustness, accuracy, and interpretability  of diverse visual localization tasks such as visual odometry, object localization, and floor plan localization.

Mohit Gupta is an Associate Professor of Computer Sciences at the University of Wisconsin Madison. He received a Ph.D. from the Robotics Institute, Carnegie Mellon University, and was a postdoctoral research scientist at Columbia University. He directs the WISION Lab with research interests in computer vision and computational imaging. He has received best paper honorable mention awards at computer vision and photography conferences in 2014 and 2019 including a Marr Prize honorable mention at IEEE ICCV, a Sony Faculty Innovation Award and an NSF CAREER award. His research is supported by NSF, ONR, DARPA, Sony, Snap, Cruise, and Intel.

Title: Shedding Light on 3D Cameras

Abstract: The advent (and commoditization) of low-cost 3D cameras is revolutionizing many application domains, including robotics, autonomous navigation, human computer interfaces, and recently even consumer devices such as cell-phones. Most modern 3D cameras (e.g., LiDAR) are active; they consist of a light source that emits coded light into the scene, i.e., its intensity is modulated over space, and/or time. The performance of these cameras is determined by their illumination coding functions. I will talk about our work on developing a coding theory of active 3D cameras. This theory, for the first time, abstracts several seemingly different 3D camera designs into a common, geometrically intuitive space. Based on this theory, we design novel 3D cameras that achieve up to an order of magnitude higher performance as compared to the current state-of-the art. I will also briefly talk about our work toward developing `All Weather’ 3D cameras that can operate in extreme real-world conditions, including outdoors (e.g., a robot navigating outdoors in bright sunlight and poor weather), and under multi-camera interference (e.g., multiple robots navigating in a shared space such as a warehouse).