Für eine korrekte Darstellung dieser Seite benötigen Sie einen XHTML-standardkonformen Browser, der die Darstellung von CSS-Dateien zulässt.

News

You are here


 

Welcome to the Visual Computing Group of the University of Siegen. The group is headed by Prof. Margret Keuper. On this website, you can find information about the group members, our courses, publications, final thesis, and open positions.
 


Research: 

Our research focus is Computer Vision, Visual Computing and Machine Learning, specifically we have recently been working on

  • Robustness in Deep Learning
  • Learning Graph Embeddings and Graph Representations
  • Neural Architecture Search
  • Grouping Problems (in vision applications such as Image and Motion Segmentation and Multiple Object Tracking)
  • Efficient Solvers for Large Grouping Problems
  • Motion Estimation
  • Image Generation and Deep Fake Detection

We also have prior work on the segmentation in volumetric bio-medical image data.



News:

01.10.2022 Tejaswini Medi joined our group!
15.09.2022 We have two accepted NeurIPS papers
15.09.2022 We still have open PhD positions. If you have questions, write an email to margret.keuper@uni-siegen.de.
08.07.2022 We have two accepted ECCV papers
15.06.2022 Shashank Agnihotri joined our group!

Recent Publications:

  • Y. Li, D. Zhang, M. Keuper, A. Khoreva, Intra-Source Style Augmentation for Improved Domain Generalization, accepted to WACV 2023.
  • A. Saseendran, K. Skubsch, S. Falkner, M. Keuper, Trading-off Image Quality for Robustness is not necessary with Regularized Deterministic Autoencoders, accepted to NeurIPS 2022.
  • J. Grabinski, P. Gavrikov, J. Keuper, M. Keuper, Robust Models are Less Over-Confident, accepted to NeurIPS 2022. 
  • J. Lukasik, S. Jung, M. Keuper, Learning Where to Look -- Generative NAS is Surprisingly Efficient, accepted to ECCV 2022.
  • J. Grabinski, S. Jung, J. Keuper, M. Keuper, FrequencyLowCut Pooling -- Plug & Play against Catastrophic Overfitting, accepted to ECCV 2022.
  • J. Grabinski, J. Keuper, M. Keuper, Aliasing and adversarial robust generalizaiton of CNNs, Machine Learning,(2022)
  • S. Jung, M. Keuper, Learning to solve Minimum Cost Multicuts efficiently using Edge-Weighted Graph Convolutional Neural Networks,  accepted to ECML-PKDD 2022.
  • Y. Zhou, W., Xiang, C. Li, B. Wang, X. Wei, L. Zhang, M. Keuper, X., Hua, SP-VIT: Learning 2D Spatial Priors for Vision Transformers, accepted to BMVC 2022.
  • S. Jung, S. Ziegler, A. Kardoost, M. Keuper, Optimizing Edge Detection for Image Segmentation with Multicut Penalties, GCPR 2022.
  • E. Levinkov*, A. Kardoost*, B. Andres and M. Keuper (*equal contribution), "Higher-Order Multicuts for Geometric Model Fitting and Motion Segmentation," in IEEE Transactions on Pattern Analysis and Machine Intelligence, doi: 10.1109/TPAMI.2022.3148795.
  • J. Siems, L. Zimmer, A. Zela, J. Lukasik, M. Keuper and F. Hutter, NAS-Bench-301 and the Case for Surrogate Benchmarks for Neural Architecture Search, ICLR, 2022.
  • A. Saseendran, K. Skubch, S. Falkner, M. Keuper, Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders: , NeurIPS 2021.
  • K. Ho, FJ Pfreundt, J Keuper, M Keuper, Estimating the Robustness of Classification Models by the Structure of the Learned Feature-Space, The AAAI-22 Workshop on Adversarial Machine Learning and Beyond, 2022
  • S Jung, M Keuper, Internalized Biases in Fréchet Inception Distance, NeurIPS 2021 Workshop on Distribution Shifts, 2021.
  • J Geiping, J Lukasik, M Keuper, M Moeller, DARTS for Inverse Problems: a Study on Hyperparameter Sensitivity (link is external), NeurIPS workshop on Inverse Problems, 2021.

 



Bachelor and Master Theses:

We are new at the University of Siegen, so we still have capacities to supervise Bachelor and Master Theses. Note that prior knowledge in our fields of research (e.g. gained by attending the lectures on Digital Image Processing, Deep Learning, etc.) is a requirement to be able to write a final thesis with us. 

Please refer to the "For Students" tab for more information.