no code implementations • 14 Feb 2024 • Edgar Heinert, Matthias Rottmann, Kira Maag, Karsten Kahl
While most of the previous works in the literature focus on the task of image classification, we go beyond this and study the texture bias of CNNs in semantic segmentation.
no code implementations • 5 Oct 2023 • Lorenc Kapllani, Long Teng, Matthias Rottmann
In this work, we study uncertainty quantification (UQ) for a class of deep learning-based BSDE schemes.
no code implementations • 30 Sep 2023 • Marius Schubert, Tobias Riedlinger, Karsten Kahl, Matthias Rottmann
Here, we propose a composite active learning framework including a label review module for deep object detection.
no code implementations • 16 Aug 2023 • Julian Burghoff, Matthias Rottmann, Jill von Conta, Sebastian Schoenen, Andreas Witte, Hanno Gottschalk
In this work, we develop a neural architecture search algorithm, termed Resbuilder, that develops ResNet architectures from scratch that achieve high accuracy at moderate computational cost.
1 code implementation • 13 Jun 2023 • Tobias Riedlinger, Marius Schubert, Sarina Penquitt, Jan-Marcel Kezmann, Pascal Colling, Karsten Kahl, Lutz Roese-Koerner, Michael Arnold, Urs Zimmermann, Matthias Rottmann
In order to address these two issues, we propose LidarMetaDetect (LMD), a light-weight post-processing scheme for prediction quality estimation.
no code implementations • 13 Mar 2023 • Marius Schubert, Tobias Riedlinger, Karsten Kahl, Daniel Kröll, Sebastian Schoenen, Siniša Šegvić, Matthias Rottmann
In this work, we for the first time introduce a benchmark for label error detection methods on object detection datasets as well as a label error detection method and a number of baselines.
no code implementations • 29 Dec 2022 • Krzysztof Lis, Matthias Rottmann, Sina Honari, Pascal Fua, Mathieu Salzmann
In other words, vision transformers trained to segment a fixed set of object classes generalize to objects well beyond this set.
no code implementations • 21 Dec 2022 • Tobias Riedlinger, Marius Schubert, Karsten Kahl, Hanno Gottschalk, Matthias Rottmann
Active learning as a paradigm in deep learning is especially important in applications involving intricate perception tasks such as object detection where labels are difficult and expensive to acquire.
no code implementations • 10 Nov 2022 • Antonia van Betteray, Matthias Rottmann, Karsten Kahl
The complexity of the weight count can be seen as a function of the number of channels, the spatial extent of the input and the number of layers of the network.
no code implementations • 18 Aug 2022 • Annika Mütze, Matthias Rottmann, Hanno Gottschalk
The main contributions of this work are 1) a modular semi-supervised domain adaptation method for semantic segmentation by training a downstream task aware CycleGAN while refraining from adapting the synthetic semantic segmentation expert 2) the demonstration that the method is applicable to complex domain adaptation tasks and 3) a less biased domain gap analysis by using from scratch networks.
1 code implementation • 13 Jul 2022 • Matthias Rottmann, Marco Reese
In this work, we for the first time present a method for detecting label errors in image datasets with semantic segmentation, i. e., pixel-wise class labels.
1 code implementation • 7 Jul 2022 • Kira Maag, Matthias Rottmann
In this work, we enhance semantic segmentation predictions using monocular depth estimation to improve segmentation by reducing the occurrence of non-detected objects in presence of domain shift.
no code implementations • 9 Jun 2022 • Robin Chan, Radin Dardashti, Meike Osinski, Matthias Rottmann, Dominik Brüggemann, Cilia Rücker, Peter Schlicht, Fabian Hüger, Nikol Rummel, Hanno Gottschalk
Finally, we include comments from industry leaders in the field of AI safety on the applicability of survey based elements in the design of AI functionalities in automated driving.
no code implementations • 30 May 2022 • Julian Burghoff, Robin Chan, Hanno Gottschalk, Annika Muetze, Tobias Riedlinger, Matthias Rottmann, Marius Schubert
Training deep neural networks is already resource demanding and so is also their uncertainty quantification.
no code implementations • 17 Feb 2022 • Robin Chan, Svenja Uhlemeyer, Matthias Rottmann, Hanno Gottschalk
However, this is in contrast to the open world assumption in automated driving that DNNs are deployed to.
1 code implementation • 31 Jan 2022 • Philipp Oberdiek, Gernot A. Fink, Matthias Rottmann
We present an approach to quantifying both aleatoric and epistemic uncertainty for deep neural networks in image classification, based on generative adversarial networks (GANs).
Out of Distribution (OOD) Detection Uncertainty Quantification
1 code implementation • 4 Jan 2022 • Svenja Uhlemeyer, Matthias Rottmann, Hanno Gottschalk
More precisely, the connected components of a predicted semantic segmentation are assessed by a segmentation quality estimate.
no code implementations • 9 Dec 2021 • Hanno Gottschalk, Matthias Rottmann, Maida Saltagic
While automated driving is often advertised with better-than-human driving performance, this work reviews that it is nearly impossible to provide direct statistical evidence on the system level that this is actually the case.
no code implementations • 29 Oct 2021 • Pascal Colling, Matthias Rottmann, Lutz Roese-Koerner, Hanno Gottschalk
We present a novel post-processing tool for semantic segmentation of LiDAR point cloud data, called LidarMetaSeg, which estimates the prediction quality segmentwise.
no code implementations • 20 Sep 2021 • Claudia Drygala, Matthias Rottmann, Hanno Gottschalk, Klaus Friedrichs, Thomas Kurbiel
In this work, we compare different statistical methods from the field of image segmentation to approach the problem of background-foreground segmentation in camera based interior sensing.
1 code implementation • 9 Jul 2021 • Tobias Riedlinger, Matthias Rottmann, Marius Schubert, Hanno Gottschalk
The vast majority of uncertainty quantification methods for deep object detectors such as variational inference are based on the network output.
no code implementations • 10 Jun 2021 • Julia Rosenzweig, Eduardo Brito, Hans-Ulrich Kobialka, Maram Akila, Nico M. Schmidt, Peter Schlicht, Jan David Schneider, Fabian Hüger, Matthias Rottmann, Sebastian Houben, Tim Wirtz
We propose a novel framework consisting of a generative label-to-image synthesis model together with different transferability measures to inspect to what extent we can transfer testing results of semantic segmentation models from synthetic data to equivalent real-life data.
2 code implementations • 30 Apr 2021 • Robin Chan, Krzysztof Lis, Svenja Uhlemeyer, Hermann Blum, Sina Honari, Roland Siegwart, Pascal Fua, Mathieu Salzmann, Matthias Rottmann
State-of-the-art semantic or instance segmentation deep neural networks (DNNs) are usually trained on a closed set of semantic classes.
no code implementations • 29 Apr 2021 • Sebastian Houben, Stephanie Abrecht, Maram Akila, Andreas Bär, Felix Brockherde, Patrick Feifel, Tim Fingscheidt, Sujan Sai Gannamaneni, Seyed Eghbal Ghobadi, Ahmed Hammam, Anselm Haselhoff, Felix Hauser, Christian Heinzemann, Marco Hoffmann, Nikhil Kapoor, Falk Kappel, Marvin Klingner, Jan Kronenberger, Fabian Küppers, Jonas Löhdefink, Michael Mlynarski, Michael Mock, Firas Mualla, Svetlana Pavlitskaya, Maximilian Poretschkin, Alexander Pohl, Varun Ravi-Kumar, Julia Rosenzweig, Matthias Rottmann, Stefan Rüping, Timo Sämann, Jan David Schneider, Elena Schulz, Gesina Schwalbe, Joachim Sicking, Toshika Srivastava, Serin Varghese, Michael Weber, Sebastian Wirkert, Tim Wirtz, Matthias Woehrle
Our paper addresses both machine learning experts and safety engineers: The former ones might profit from the broad range of machine learning topics covered and discussions on limitations of recent methods.
1 code implementation • 14 Dec 2020 • Kira Maag, Matthias Rottmann, Serin Varghese, Fabian Hueger, Peter Schlicht, Hanno Gottschalk
In this paper, we present a time-dynamic approach to model uncertainties of instance segmentation networks and apply this to the detection of false positives as well as the estimation of prediction quality.
1 code implementation • ICCV 2021 • Robin Chan, Matthias Rottmann, Hanno Gottschalk
In our experiments we consistently observe a clear additional gain in OoD detection performance, cutting down the number of detection errors by up to 52% when comparing the best baseline with our results.
no code implementations • 24 Nov 2020 • Hayk Asatryan, Hanno Gottschalk, Marieke Lippert, Matthias Rottmann
In recent years, generative adversarial networks (GANs) have demonstrated impressive experimental results while there are only a few works that foster statistical learning theory for GANs.
no code implementations • 7 Oct 2020 • Kamil Kowol, Matthias Rottmann, Stefan Bracke, Hanno Gottschalk
In this work, we present an uncertainty-based method for sensor fusion with camera and radar data.
no code implementations • 5 Oct 2020 • Pascal Colling, Lutz Roese-Koerner, Hanno Gottschalk, Matthias Rottmann
The comparison and analysis of the results provide insights into annotation costs as well as robustness and variance of the methods.
1 code implementation • 4 Oct 2020 • Marius Schubert, Karsten Kahl, Matthias Rottmann
On the other hand, meta regression gives rise to a quality estimate.
no code implementations • 23 Sep 2020 • Matthias Rottmann, Kira Maag, Mathis Peyron, Natasa Krejic, Hanno Gottschalk
In this work we outline a mathematical proof that the CW attack can be used as a detector itself.
1 code implementation • 14 May 2020 • Philipp Oberdiek, Matthias Rottmann, Gernot A. Fink
When deploying deep learning technology in self-driving cars, deep neural networks are constantly exposed to domain shifts.
no code implementations • 16 Dec 2019 • Robin Chan, Matthias Rottmann, Fabian Hüger, Peter Schlicht, Hanno Gottschalk
We present proof-of-concept results for CIFAR-10, and prove the efficiency of our method for the semantic segmentation of street scenes on the Cityscapes dataset based on predicted instances of the 'human' class.
1 code implementation • 8 Dec 2019 • Matthias Rottmann, Kira Maag, Robin Chan, Fabian Hüger, Peter Schlicht, Hanno Gottschalk
In recent years, deep learning methods have outperformed other methods in image recognition.
1 code implementation • 12 Nov 2019 • Kira Maag, Matthias Rottmann, Hanno Gottschalk
In the semantic segmentation of street scenes with neural networks, the reliability of predictions is of highest interest.
no code implementations • 2 Jul 2019 • Robin Chan, Matthias Rottmann, Radin Dardashti, Fabian Hüger, Peter Schlicht, Hanno Gottschalk
Neural networks for semantic segmentation can be seen as statistical models that provide for each pixel of one image a probability distribution on predefined classes.
1 code implementation • 9 Apr 2019 • Matthias Rottmann, Marius Schubert
For the task of meta classification we obtain a classification accuracy of $81. 93\%$ and an AUROC of $89. 89\%$.
1 code implementation • 24 Jan 2019 • Robin Chan, Matthias Rottmann, Fabian Hüger, Peter Schlicht, Hanno Gottschalk
We approach such potential misclassifications by weighting the posterior class probabilities with the prior class probabilities which in our case are the inverse frequencies of the corresponding classes in the training dataset.
1 code implementation • 1 Nov 2018 • Matthias Rottmann, Pascal Colling, Thomas-Paul Hack, Robin Chan, Fabian Hüger, Peter Schlicht, Hanno Gottschalk
We aggregate these dispersion measures segment-wise and derive metrics that are well-correlated with the segment-wise IoU of prediction and ground truth.
no code implementations • 22 May 2018 • Philipp Oberdiek, Matthias Rottmann, Hanno Gottschalk
If we however allow the meta classifier to be trained on uncertainty metrics for some out-of-distribution samples, meta classification for concepts remote from EMNIST digits (then termed known unknowns) can be improved considerably.
1 code implementation • 3 Mar 2018 • Matthias Rottmann, Karsten Kahl, Hanno Gottschalk
In a setting where a small amount of labeled data as well as a large amount of unlabeled data is available, our method first learns the labeled data set.