Postdoctoral Researcher
Cornell-Tech
PhD '19 Tel Aviv University
Areas of Interest
- Computer Vision
- Computer Graphics
Poster
Controlling Visual Data Generation
Abstract
Imagery is used everywhere to understand and communicate our world. However, while images are encoded as collections of raw, low-level 2D pixels, we would like to understand and manipulate them at the higher-level of objects and concepts. Deep neural networks have proven an effective way for machines to model images, embedding the low-level image modality onto a latent high dimensional space. However, these deep models are very difficult for humans to understand and control. Such control is critical in many practical applications, like an artist who wants to use such models as an aid to (but not a full replacement for) their work. My research leverages underlying structure to adapt how we model and manipulate visual data. By combining pixels with more structured modalities, including text, documents and 3D geometry, I develop computer vision algorithms and computer graphics techniques and interfaces that are better suited for handling the full complexity of the visual world.
Bio
Hadar Averbuch-Elor is a postdoctoral researcher at Cornell-Tech working with Prof. Noah Snavely. She completed her PhD in Electrical Engineering at Tel-Aviv University in Israel where she was advised by Prof. Daniel Cohen-Or. She also held research positions at Facebook and Amazon AI. Her research focuses on modeling and manipulating visual concepts by combining pixels with more structured modalities, including natural language and 3D geometry. Hadar is supported by the Zuckerman STEM Leadership Program and the Schmidt Futures Program.
