publications

2023

1. A Call to Reflect on Evaluation Practices for Failure Detection in Image Classification

   Jaeger, Paul F,  Lüth, Carsten T., Klein, Lukas, and Bungert, Till J.

   ICLR 2023 Jan 2023

   HTML Code
2. CRADL: Contrastive Representations for Unsupervised Anomaly Detection and Localization

   Lüth, Carsten T., Zimmerer, David, Koehler, Gregor, Jaeger, Paul F., Isensee, Fabian, Petersen, Jens, and Maier-Hein, Klaus H.

   arXiv:2301.02126 Jan 2023

   HTML
3. Contrastive Representations for Unsupervised Anomaly Detection and Localization

   Lüth, Carsten T., Zimmerer, David, Koehler, Gregor, Jaeger, Paul F., Isenensee, Fabian, and Maier-Hein, Klaus H.

   Bildverarbeitung für die Medizin 2023 Jun 2023

   HTML
4. cOOpD: Reformulating COPD Classification on Chest CT Scans as Anomaly Detection Using Contrastive Representations

   Almeida, Silvia D,  Lüth, Carsten T, Norajitra, Tobias, Wald, Tassilo, Nolden, Marco, Jäger, Paul F, Heussel, Claus P, Biederer, Jürgen, Weinheimer, Oliver, and Maier-Hein, Klaus H

   In MICCAI 2023 Oct 2023

   HTML Code
5. Navigating the pitfalls of active learning evaluation: A systematic framework for meaningful performance assessment

   Lüth, Carsten T, Bungert, Till J, Klein, Lukas, and Jaeger, Paul F

   In NeurIPS 2023 Dez 2023

2022

1. Unsupervised Anomaly Detection in the Wild

   Zimmerer, David, Paech, Daniel,  Lüth, Carsten, Petersen, Jens, Köhler, Gregor, and Maier-Hein, Klaus

   Dez 2022

   HTML
2. Probabilistic Modeling of COPD Imaging Characteristics for Disease Severity Prediction

   Dias Almeida, S, Norajitra, T,  Lüth, C, Wald, T, Kopytova, T, Nolden, M, F. Jaeger, P, Peter Heussel, C, Biederer, J, Kauczor, H, Weinheimer, O, and Maier-Hein, K

   Sep 2022

   HTML

2019

1. Guided Image Generation with Conditional Invertible Neural Networks

   Ardizzone, Lynton,  Lüth, Carsten, Kruse, Jakob, Rother, Carsten, and Köthe, Ullrich

   arXiv:1907.02392 [cs] Jul 2019

   Abs HTML

   In this work, we address the task of natural image generation guided by a conditioning input. We introduce a new architecture called conditional invertible neural network (cINN). The cINN combines the purely generative INN model with an unconstrained feed-forward network, which efficiently preprocesses the conditioning input into useful features. All parameters of the cINN are jointly optimized with a stable, maximum likelihood-based training procedure. By construction, the cINN does not experience mode collapse and generates diverse samples, in contrast to e.g. cGANs. At the same time our model produces sharp images since no reconstruction loss is required, in contrast to e.g. VAEs. We demonstrate these properties for the tasks of MNIST digit generation and image colorization. Furthermore, we take advantage of our bi-directional cINN architecture to explore and manipulate emergent properties of the latent space, such as changing the image style in an intuitive way.