Ming-Yu Liu is a researcher at Nvidia Research, working on computer vision and machine learning. Prior to joining NVIDIA, he was a researcher at Mitsubishi Electric Research Labs (MERL). He received his Ph.D. from the Department of Electrical and Computer Engineering at the University of Maryland College Park in 2012. His early research work on object pose estimation contributed to development of the first commercial vision-based robotic bin-picking system for robotic assembly tasks, which was awarded the 100 most innovative technology products of the year by the R&D magazine in 2014. He is a recipient of a best paper finalist award from the robotics science and system (RSS) 2015 conference for his street scene understanding work. Recently, his research focus shifted to deep generative models for image understanding and generation. His goal is to enable machines superhuman-like imagination capabilities. He has 10+ US patents and 20+ papers on firs tier computer vision, machine learning, and robotics conferences.

Check out my new code release in GitHub. Follow me on Twitter Twitter


Publication

    2015

    • Layered Interpretation of Street View Images
      Ming-Yu Liu, Shuoxin Lin, Srikumar Ramalingam, Oncel Tuzel
      Robotics: Science and Systems Conference (RSS) Best paper finalist, 2015, Rome, Italy
      arXiv preprint arXiv:1506.04723

    2011


Patents granted

  • US 9,558,268: Method for Semantically Labeling an Image of a Scene using Recursive Context Propagation

  • US 9,704,257: System and method for semantic segmentation using Gaussian random field network

  • US 9,633,274: Method and system for denoising images using deep Gaussian conditional random field network

  • US 9,558,268: Method for semantically labeling an image of a scene using recursive context propagation

  • US 8,428,363: Method for segmenting images using superpixels and entropy rate clustering

  • US 8,983,177: Method for increasing resolutions of depth images

  • US 8,908,913: Voting-based pose estimation for 3D sensors

  • US 9,195,904: Method for detecting objects in stereo images

  • US 9,280,827: Method for determining object poses using Weighted Features

  • My MERL patents


Education

  • PhD, Electrical&Computer Engineering, University of Maryland College Park, Advisor: Rama Chellappa, 2006-2012


Tutorial

  • CVPR2017: Theory and Applications of Generative Adversarial Networks, [Site]

  • ACCV2016: Deep Learning for Vision Guided Language Generation and Image Generation, [Site]


Awards


Research


Public Service

  • Reviewers: NIPS, CVPR, ECCV, ICCV, IEEE TPAMI, IEEE TIP, IEEE SPL, CVIU,

  • Area Chairs: WACV

  • NSF proposal reviewer