Abstract
Building realistic and accurate scene models of the world has long remained a central goal of shape analysis. While it is now easy to capture large volumes of data including images, videos, scans, converting such data to a factorized representation remain a major challenge. For example, given an image, can we tell what are the objects in the scene, how are they illuminated, or how will they behave in presence of external forces. In this talk, I will discuss some of our recent attempts to factorize raw measurements into scene geometry, appearance, and their interactions. I will discuss how synthetically rendered images can be used to discover object arrangements in photographs, capture real world illumination and texture using geometric proxies, and ‘read off’ physical object properties by observing them collide in space. Our methods allow for large-scale unsupervised production of richly textured 3D models directly from image data, providing high quality realistic objects for 3D scene design or photo editing applications, as well as a wealth of data for training machine learning algorithms for various inference tasks in graphics and vision.
Speaker Bio
Prof. Niloy J. Mitra leads the Smart Geometry Processing group in the Department of Computer Science at University College London. He received his PhD degree from Stanford University under the guidance of Leonidas Guibas. His research interests include shape analysis, computational design and fabrication, and geometry processing. Niloy received the ACM Siggraph Significant New Researcher Award in 2013 and the BCS Roger Needham award in 2015. His work has twice been selected and featured as research highlights in the Communication of ACM, received best paper award at ACM Symposium on Geometry Processing 2014, and Honourable Mention at Eurographics 2014. He is on the Editorial Board of ACM Transactions on Graphics. Besides research, Niloy is an active DIYer and loves reading, bouldering, and cooking.