Lecture Series “Robust AI”: Annika Mütze

The next lecture in our lecture series will be given by Dr. Annika Mütze (University of Wuppertal) on the topic of “Influence of Texture, Shape and Color for Learning Semantic Segmentation”. This talk ist hosted by Dr. Kaja Balzereit, Hochschule Bielefeld.

When & where:

Thursday, February 14, 2025, 4:15pm – 5:45pm at Hochschule Bielefeld (HSBI), Room E2 or online via Zoom (https://hsbi-de.zoom-x.de/j/67883725986?pwd=zkor8TEOG6NPlVRBOSVvtKaUwmkcIw.1).

Abstract:

In recent years, a body of works has emerged, studying shape and texture biases of off-the-shelf pre-trained deep neural networks (DNN) for image classification. These works study how much a trained DNN relies on image cues, predominantly shape and texture. In this talk, we switch the perspective, posing the following questions: What can a DNN learn from each of the image cues, i.e., shape, texture and color, respectively? How much does each cue influence the learning success? And what are the synergy effects between different cues? 

We study these questions on semantic segmentation which allows us to address our questions on pixel level. To conduct this I introduce our generic procedure to decompose a given dataset into multiple ones, each of them only containing either a single cue or a chosen mixture. These datasets are then used to learn cue experts which allow us to study the influences of the different cues on the learning task.

At the end of the talk, we will come back to the bias perspective by giving an outlook on how the cue datasets can help to analyze cue biases in the context of semantic segmentation.

Speaker:

Annika Mütze is postdoctoral researcher at the ACM AI Lab (AI Lab of the Applied and Computational Mathematics Group) at the University of Wuppertal. In her research, she focuses on Safe AI and aims at a deeper understanding on how deep neural networks learn, where they fail and how to make them more robust.
Prior to that, she was a research assistant at the Fraunhofer Institute for Intelligent Analysis and Information Systems (IAIS) in Sankt Augustin and did her PhD on “Visual Recognition Using Deep Neural Networks on Abstract and Decomposed Data” supervised by Prof. Dr. Hanno Gottschalk and PD Dr. Matthias Rottmann in the Stochastics Group at the University of Wuppertal.