2022 Max Planck Lecture & Intelligent Systems Summer Colloquium

K.K. Lee Professor of Chemical Engineering,
and by courtesy Professor of Chemistry,
Materials Science and Engineering Department Chair,
Department of Chemical Engineering Director,
Stanford Wearable Electronics Initiative (eWEAR)
Stanford University

Detailed information about the Max Planck Lecture can be found on our Lecture page.

Wieland Brendel received his Diploma in physics from the University of Regensburg (2010) and his Ph.D. in computational neuroscience from the École normale supérieure in Paris (2014). He joined the University of Tübingen as a postdoctoral researcher in the group of Matthias Bethge, became a Principal Investigator and Team Lead in the Tübingen AI Center (2018) and an Emmy Noether Group Leader for Robust Machine Learning (2020). In May 2022, Wieland joined the Max-Planck Institute for Intelligent Systems as an independent Group Leader. Besides working on robust, generalisable and interpretable machine vision, he co-founded a nationwide school competition (bw-ki.de) and a machine learning startup focused on visual quality control (layer7.ai).

Machines perceive the world very differently from humans. This gap between humans and machines has narrowed little in the past years, despite even the recent emergence of impressive large-scale foundation models. In this talk, I discuss our recent systematic work to unravel the hidden assumptions underlying existing representation learning techniques and to develop new techniques that leverage the compositional 3D nature of our visual world in a more principled way.

Max Planck Research Group Leader,
Max Planck Institute for Intelligent Systems,
Tübingen & Tübingen AI Center

A challenge in data science is modeling 3-dimensional shapes such as a brain tumor or a bone. We introduce a representation of shapes that allows us to model shapes without requiring landmarks or assuming the shapes are diffeomorphic, we can also model shapes that are topologically different. We show that these transforms are injective on the space of shapes — each shape has a unique transform. We also provide a bound on the complexity of a  moduli space of shapes (characterized by bounds on critical points and curvature). We also introduce to date the most general shape space that has a mathematically rigorous definition, the construction is sheaf theoretic. We apply this representation to four statistical challenges: (1) computing the similarity between 3-dimensional shapes; (2) using the 3-dimensional shape as a covariate in regression problems; and (3) sub-image selection or what parts of the shape is associated to variation in a phenotype or trait, and (4) learning formal evolutionary models for shapes, that is estimating parameters like the selection pressure.

• Alexander von Humboldt Professor
• University Leipzig Center for Scalable Data Analytics and Artificial Intelligence
• Max Planck Institute for Mathematics in the Sciences

Betty Mohler has over 100 research articles published on space and body perception, virtual humans, and virtual reality. Betty joined Amazon in 2018 as a Principal Research Scientist at the Amazon Research Development Center. Previously, she was a Professor at the Cognitive Science Center at Technical University of Darmstadt, and for ten years she was an Independent Research Group Leader of "Space and Body Research Group" at the Max Planck Institute for Biological Cybernetics. At TU Darmstadt, the Max Planck Institutes and at the University of Utah (where Betty received her PhD in computer science), she had the privilege of working with world-renowned experts in multi-disciplinary teams focused on Computer Graphics, Virtual Reality, Computer Vision, Philosophy of Neuroscience, and Neural Behavioral Science. Betty and her research teams have always focused equally on human perception, as well as developing immersive technology. Her passion for developing virtual reality for medical, clinical and industrial applications, along with a central focus on the person led her to a primary research focus of "Self-avatars in Immersive Technology".

Betty Mohler joined Amazon in 2018 as a Principal Research Scientist at the Amazon Research Development Center. However, the research of this presentation was conducted prior to Betty Mohler’s position at Amazon and therefore does not reflect her research or future work at Amazon. Betty spent 15 years in academia (largely at the Max Planck Institute for Biological Cybernetics) working on head-mounted and large-screen immersive virtual reality as well as self-avatar tools. She studied human spatial and body perception with the aim of investigating multi-sensory perception, action and social interactions. In parallel Betty & her team developed better technical tools for scientists and neural and psychological clinics/hospitals. In this talk she will share results and findings of several studies on self-body size perception, stylized characters and attitudes towards bodies. Further she’ll reflect on body positivity and the language used to discuss bodies (both digital and real).

Betty Mohler has a Ph.D. and her affiliation is with Amazon Development Center Germany GmbH