Generative Representations of the World

Speaker: Dr Ilke Demir , DeepScale

Title: Generative Representations of the World

Abstract: Generative approaches enable creating numerous scenarios coherent with the reality, as long as the representations are good approximations of the real world. In this talk, I will discuss extracting such generative representations from 2D and 3D data for mapping, modeling, and reconstruction of urban models and spatial data; combining computer vision, machine learning, and computational geometry approaches. I will introduce geometry processing algorithms to exploit similarities, grammar discovery approaches to extract procedural rules, and machine learning methods to understand geospatial information. The talk will conclude by proposed applications, new directions in 3D deep learning for generative models, and experimental results on various types of geometric data.

Bio: Ilke Demir earned her PhD Degree in Computer Science from Purdue University in 2016, focusing on 3D vision approaches for generative models, urban reconstruction and modeling, and computational geometry for synthesis and fabrication. Afterwards, Dr. Demir joined Facebook as a Postdoctoral Research Scientist working with Ramesh Raskar from MIT. Her research included human behavior analysis via deep learning in virtual reality, geospatial machine learning, and 3D reconstruction at scale. In addition to her publications in top-tier venues (SIGGRAPH, ICCV, CVPR), she has organized workshops, competitions, and courses in the intersection of deep learning and computer vision. She has received several awards and honors such as Jack Dangermond Award, Bilsland Dissertation Fellowship, and Industry Distinguished Lecturer, in addition to her best paper/poster/reviewer awards. Currently she builds efficient deep learning architectures for autonomous driving tasks as a Senior Research Scientist at DeepScale.