@unpublished{WolfBraunUlrich2024, author = {Wolf, Dominik Werner and Braun, Alexander and Ulrich, Markus}, title = {Optical aberrations in autonomous driving: Physics-informed parameterized temperature scaling for neural network uncertainty calibration}, series = {International Journal of Computer Vision (IJCV)}, journal = {International Journal of Computer Vision (IJCV)}, publisher = {arXiv}, doi = {10.48550/arXiv.2412.13695}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-50120}, year = {2024}, abstract = {'A trustworthy representation of uncertainty is desirable and should be considered as a key feature of any machine learning method' (Huellermeier and Waegeman, 2021). This conclusion of Huellermeier et al. underpins the importance of calibrated uncertainties. Since AI-based algorithms are heavily impacted by dataset shifts, the automotive industry needs to safeguard its system against all possible contingencies. One important but often neglected dataset shift is caused by optical aberrations induced by the windshield. For the verification of the perception system performance, requirements on the AI performance need to be translated into optical metrics by a bijective mapping (Braun, 2023). Given this bijective mapping it is evident that the optical system characteristics add additional information about the magnitude of the dataset shift. As a consequence, we propose to incorporate a physical inductive bias into the neural network calibration architecture to enhance the robustness and the trustworthiness of the AI target application, which we demonstrate by using a semantic segmentation task as an example. By utilizing the Zernike coefficient vector of the optical system as a physical prior we can significantly reduce the mean expected calibration error in case of optical aberrations. As a result, we pave the way for a trustworthy uncertainty representation and for a holistic verification strategy of the perception chain.}, language = {en} } @unpublished{WolfBalajiBraunetal.2024, author = {Wolf, Dominik Werner and Balaji, Prasannavenkatesh and Braun, Alexander and Ulrich, Markus}, title = {Decoupling of neural network calibration measures}, series = {German Conference on Pattern Recognition (GCPR) 2024}, journal = {German Conference on Pattern Recognition (GCPR) 2024}, publisher = {arXiv}, doi = {10.48550/arXiv.2406.02411}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-50104}, year = {2024}, abstract = {A lot of effort is currently invested in safeguarding autonomous driving systems, which heavily rely on deep neural networks for computer vision. We investigate the coupling of different neural network calibration measures with a special focus on the Area Under the Sparsification Error curve (AUSE) metric. We elaborate on the well-known inconsistency in determining optimal calibration using the Expected Calibration Error (ECE) and we demonstrate similar issues for the AUSE, the Uncertainty Calibration Score (UCS), as well as the Uncertainty Calibration Error (UCE). We conclude that the current methodologies leave a degree of freedom, which prevents a unique model calibration for the homologation of safety-critical functionalities. Furthermore, we propose the AUSE as an indirect measure for the residual uncertainty, which is irreducible for a fixed network architecture and is driven by the stochasticity in the underlying data generation process (aleatoric contribution) as well as the limitation in the hypothesis space (epistemic contribution).}, language = {en} } @article{JakabBraunAgnewetal.2024, author = {Jakab, Daniel and Braun, Alexander and Agnew, Cathaoir and Mohandas, Reenu and Deegan, Brian Michael and Molloy, Dara and Ward, Enda and Scanlan, Anthony and Eising, Ciar{\´a}n}, title = {SS-SFR: synthetic scenes spatial frequency response on Virtual KITTI and degraded automotive simulations for object detection}, series = {IET Conference Proceedings}, volume = {2024}, journal = {IET Conference Proceedings}, number = {10}, publisher = {Institution of Engineering and Technology (IET)}, issn = {2732-4494}, doi = {10.1049/icp.2024.3292}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-50080}, pages = {110 -- 117}, year = {2024}, language = {en} } @article{WolfThielbeerUlrichetal.2024, author = {Wolf, Dominik Werner and Thielbeer, Boris and Ulrich, Markus and Braun, Alexander}, title = {Wavefront aberration measurements based on the Background Oriented Schlieren method}, series = {Measurement: Sensors}, journal = {Measurement: Sensors}, publisher = {Elsevier}, issn = {2665-9174}, doi = {10.1016/j.measen.2024.101509}, year = {2024}, abstract = {Applications based on neural networks tend to be very sensitive to dataset shifts. Hence, the perception chain for autonomous driving is safeguarded against perturbations by imposing exaggerated optical quality requirements. Due to the non-linear coupling of optical elements, system requirements of camera-based Advanced Driver Assistance Systems (ADAS) can not be easily decomposed into individual part tolerances of the objective lens and the windscreen. This holds true for intensity-based part measurements, e.g. slanted edge measurements according to ISO12233, which can fundamentally not capture interference effects of the complex light field. Instead wavefront-based part measurements are required. Unfortunately, state-of-the-art wavefront measurement techniques are limited by the spanned sensitive area of the Shack-Hartmann lenslet array or the aperture stop of a corresponding interferometrical setup, respectively. Further, both measurements are limited by using collimated light only, whereas the target application has a large (angular) field of view, requiring many different measurements. We address those bottlenecks by proposing a novel wavefront aberration measurement procedure based on the Background Oriented Schlieren (BOS) method utilizing image auto-correlation. We analytically derive the governing equations for determining the Zernike coefficients of a wavefront aberration map in the knowledge of the local refractive power map obtained by a high-resolution BOS measurement. Furthermore, we experimentally demonstrate the feasibility of the measurement technique. Applying this novel method yields the promise of affordable wavefront aberration measurements only requiring a high-resolution camera and a sophisticated alignment strategy.}, language = {en} } @article{MolloyMuellerDeeganetal.2024, author = {Molloy, Dara and M{\"u}ller, Patrick and Deegan, Brian and Mullins, Darragh and Horgan, Jonathan and Ward, Enda and Jones, Edward and Braun, Alexander and Glavin, Martin}, title = {Analysis of the Impact of Lens Blur on Safety-Critical Automotive Object Detection}, series = {IEEE Access}, volume = {12}, journal = {IEEE Access}, publisher = {IEEE}, issn = {2169-3536}, doi = {10.1109/ACCESS.2023.3348663}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-42999}, pages = {3554 -- 3569}, year = {2024}, abstract = {Camera-based object detection is widely used in safety-critical applications such as advanced driver assistance systems (ADAS) and autonomous vehicle research. Road infrastructure has been designed for human vision, so computer vision, with RGB cameras, is a vital source of semantic information from the environment. Sensors, such as LIDAR and RADAR, are also often utilized for these applications; however, cameras provide a higher spatial resolution and color information. The spatial frequency response (SFR), or sharpness of a camera, utilized in object detection systems must be sufficient to allow a detection algorithm to localize objects in the environment over its lifetime reliably. This study explores the relationship between object detection performance and SFR. Six state-of-the-art object detection models are evaluated with varying levels of lens defocus. A novel raw image dataset is created and utilized, containing pedestrians and cars over a range of distances up to 100-m from the sensor. Object detection performance for each defocused dataset is analyzed over a range of distances to determine the minimum SFR necessary in each case. Results show that the relationship between object detection performance and lens blur is much more complex than previous studies have found due to lens field curvature, chromatic aberration, and astigmatisms. We have found that smaller objects are disproportionately impacted by lens blur, and different object detection models have differing levels of robustness to lens blur}, language = {en} } @article{GeerkensSieberichsBraunetal.2024, author = {Geerkens, Simon and Sieberichs, Christian and Braun, Alexander and Waschulzik, Thomas}, title = {QI²: an interactive tool for data quality assurance}, series = {AI and Ethics}, volume = {4}, journal = {AI and Ethics}, publisher = {Springer Nature}, issn = {2730-5961}, doi = {10.1007/s43681-023-00390-6}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-42945}, pages = {141 -- 149}, year = {2024}, abstract = {The importance of high data quality is increasing with the growing impact and distribution of ML systems and big data. Also, the planned AI Act from the European commission defines challenging legal requirements for data quality especially for the market introduction of safety relevant ML systems. In this paper, we introduce a novel approach that supports the data quality assurance process of multiple data quality aspects. This approach enables the verification of quantitative data quality requirements. The concept and benefits are introduced and explained on small example data sets. How the method is applied is demonstrated on the well-known MNIST data set based an handwritten digits.}, language = {en} } @article{SieberichsGeerkensBraunetal.2024, author = {Sieberichs, Christian and Geerkens, Simon and Braun, Alexander and Waschulzik, Thomas}, title = {ECS: an interactive tool for data quality assurance}, series = {AI and Ethics}, volume = {4}, journal = {AI and Ethics}, publisher = {Springer Nature}, issn = {2730-5961}, doi = {10.1007/s43681-023-00393-3}, url = {http://nbn-resolving.de/urn:nbn:de:hbz:due62-opus-42908}, pages = {131 -- 139}, year = {2024}, abstract = {With the increasing capabilities of machine learning systems and their potential use in safety-critical systems, ensuring high-quality data is becoming increasingly important. In this paper, we present a novel approach for the assurance of data quality. For this purpose, the mathematical basics are first discussed and the approach is presented using multiple examples. This results in the detection of data points with potentially harmful properties for the use in safety-critical systems.}, language = {en} }