Meet Jan Eric Lenssen at our Tübingen AI Talk Series #7
We are delighted to announce the next speaker in our Tübingen AI Talk series: Jan Eric Lenssen.
Details of the talk:
- Date: May 2, 2024
- Time: 11:00 a.m. - 12:00 p.m.
- Location: Ground-floor lecture hall, Tübingen AI Center (Maria-von-Linden-Str. 6, 72076 Tübingen)
Talk title: Inferring the 3D world from incomplete observations: representation and reasoning on scenes and objects
Abstract: Computer Vision has become increasingly capable of representing the 3D world, given large sets of dense observations, i.e. images or videos. However, in comparison with humans, we still lack an important ability: deriving 3D representations from just a few, incomplete 2D observations. Replicating this ability might be the next important step towards a general vision system. A key aspect of the human abilities is that observations are complemented by previously learned information: the world is not only sensed - to a large degree it is inferred. The common way to approach this task with deep learning are data priors, which capture information present in large datasets and which are used to perform inference from novel observations.
In this talk I will present our recent advances towards this goal, consisting of novel methods to represent and reason about 3D scenes and objects. The focus will be on efficient representations and models that can be trained on images and videos, as well as on two specific aspects of reconstruction: First, we will discuss the notion of uncertainty in the context of 3D reconstruction of scenes, which we believe to be a crucial concept of 3D inference. Second, we investigate specific requirements for modeling (non-rigid) 3D objects and present a simple but powerful representation based on coherent point templates that lends itself to a variety of tasks. The presented concepts are showcased on examples from my collaborating groups.
Bio: Jan Eric Lenssen is a senior researcher and group leader at MPI for Informatics in Saarbrücken and Saarland University Associate Fellow. He obtained his PhD in 2022 from TU Dortmund University in the area of graph neural networks and geometric deep learning, before starting as a PostDoc at MPI and becoming group leader in 2023. During his PhD, he did internships at Meta Reality Labs and Nnaissense. He also is part of the founding team of kumo.ai, a Stanford/TU Dortmund startup that brings graph neural networks to relational databases.