Machine Learning and Friends Lunch (Online)

Title: "Practical 3D Object Understanding from Image Collections and Videos"

Abstract: Much of computer vision is understanding objects around us. In this talk, I will give an overview of our research works on understanding 3D object properties from 2D image collections and videos. Annotating several of the 3D object properties such as 3D shape, material properties, etc. is very labor-intensive and can not easily scale to large-scale datasets and new object categories. So, techniques that can estimate these object properties with minimal supervision are important for practical purposes.

In the first part of the talk, I will present an overview of our research works that can simultaneously learn object shape and material properties i.e., estimating re-lightable 3D assets from an image collection of an object that is captured under unconstrained variable lighting environments. In the second part, I will present techniques to estimate articulated 3D object shape given a single monocular video of a moving object. These works are category-agnostic and only assume a weak form of 2D supervision (object segmentations), thereby providing important steps towards practical 3D object understanding from the 2D world imagery.

Bio: Varun Jampani is a researcher at Google Research in Cambridge, US. Prior to that, he was a researcher at NVIDIA. He works in the areas of machine learning and computer vision and his main research interests include content-adaptive neural networks, self-supervised visual discovery and novel view synthesis. He obtained his PhD with highest honors at Max Planck Institute for Intelligent Systems (MPI) and the University of Tübingen in Tübingen, Germany. He obtained his BTech and MS from the International Institute of Information Technology, Hyderabad (IIIT-H), India, where he was a gold medalist. His work on 'SplatNet' has received 'Best Paper Honorable Mention' award at CVPR'18.

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).