Title: Neural fields beyond novel view synthesis
Abstract: Neural 3D representations have unlocked the potential to achieve 3D perception solely from 2D supervision. Yet target problem space is relatively restricted: we assume images to be equipped with near-perfect camera poses, rather than generalizing we generally overfit to a single-scene, and we evaluate results mostly from a novel-view synthesis perspective, generally ignoring how to integrate other sources of information (e.g. user annotations and/or other 3D sensing mechanisms). In this talk I will discuss how we address these shortcomings in recently published work (links below), in particular with applications to faces and environments.
Bio: Andrea Tagliasacchi is a staff research scientist at Google Brain and an adjunct faculty in the computer science department at the University of Toronto. Starting August 2022, he will join Simon Fraser University (SFU) as an Associate Professor and Visual Visual Computing Research Chair. His research focuses on 3D perception, which lies at the intersection of computer vision, computer graphics and machine learning. In 2018, he was invited to join Google Daydream as a visiting faculty and eventually joined Google full time in 2019. Before joining Google, he was an assistant professor at the University of Victoria (2015-2017), where he held the Industrial Research Chair in 3D Sensing. His alma mater include EPFL (postdoc) SFU (PhD, NSERC Alexander Graham Bell fellow) and Politecnico di Milano (MSc, gold medalist). His research has won several awards including the CVPR 2021 best student paper award, and the SGP 2015 best paper award.
To obtain the Zoom link for this event, please see the event announcements from MLFL on the college email lists or contact wenlongzhao [at] cs.umass.edu (subject: MLFL%20Zoom%20Link) (Wenlong Zhao).