arXiv subjects

Peter Bauer

Publications and source records attributed to Peter Bauer.

14 recordsLinked to original sources

arXiv2026

Machine learning is revolutionizing weather forecasting -- the next step is a change in how we work

Following the success of machine learning in producing weather predictions with competitive skill compared to complex traditional systems, this article shifts attention from forecast output to the working practices that make prediction systems possible. We argue that machine learning and recent digital technologies will reshape the forecasting value chain: how models are coded and developed, how observations and Earth-system data are exploited, how data and computing are managed, how systems are verified, and how information is created, evaluated and turned into services. We discuss six non-exhaustive areas in which agentic software engineering, open and compressed data, shared verification workflows, interactive computing and generative methods may make modelling, evaluation and service creation faster, more interactive and more widely accessible. These changes will require weather and climate centres to adapt their infrastructures, data stewardship, trust and quality-assurance frameworks, skills and service delivery while maintaining scientific understanding, operational reliability, human expertise and their public-service role.

physics.ao-ph↗

arXiv2026

PDRNN: Modular Data-driven Pedestrian Dead Reckoning on Loosely Coupled Radio- and Inertial-Signalstreams

Modern pedestrian dead reckoning (PDR) systems rely on fusing noisy and biased estimates of position, velocity, and calibrated orientation derived from loosely coupled sensors to determine the current pose of a localized object. However, discrepancies in the sampling rates of sensor-specific estimation methods and unreliable transmission pose significant challenges. And traditional methods often fail to effectively fuse multimodal sensor data during dynamic movements characterized by high accelerations, velocities, and rapidly varying orientations. To address these limitations, we propose a simple recurrent neural network (RNN) architecture capable of implicitly forecasting asynchronous sensor data streams from diverse estimation methods along reference trajectories. The proposed approach introduces PDRNN, a modular hybrid AI-assisted PDR system that handles each component as an independent ensemble of machine learning (ML) models to estimate both key parameter means and variances. Separate ML-based models are employed to estimate orientation, (un)directed velocity or distance from acceleration and gyroscope data, with optional absolute positioning from synchronized radio systems such as 5G for stabilization. A final fusion model combines these outputs, position, velocity, and orientation, while using uncertainty estimates to enhance system robustness. The modular design allows individual components to be updated, fine-tuned, or replaced without affecting the entire system. Experiments on dynamic sports movement data show that PDRNN achieves superior accuracy and precision compared to classic and ML-based methods, effectively avoiding error accumulation common in black-box approaches. And PDRNN offers forecast capabilities and better component control despite increased system complexity.

cs.LG↗

arXiv2025

Page Classification for Print Imaging Pipeline

Digital copiers and printers are widely used nowadays. One of the most important things people care about is copying or printing quality. In order to improve it, we previously came up with an SVM-based classification method to classify images with only text, only pictures or a mixture of both based on the fact that modern copiers and printers are equipped with processing pipelines designed specifically for different kinds of images. However, in some other applications, we need to distinguish more than three classes. In this paper, we develop a more advanced SVM-based classification method using four more new features to classify 5 types of images which are text, picture, mixed, receipt and highlight.

cs.CV↗

arXiv2024

What if? Numerical weather prediction at the crossroads

This paper provides an outlook on the future of operational weather prediction given the recent evolution in science, computing and machine learning. In many parts, this evolution strongly deviates from the strategy operational centres have formulated only several years ago. New opportunities in digital technology have greatly accelerated progress, and the full integration of computational science in numerical weather prediction centres is common knowledge now. Within the last few years, a vast machine learning research community has emerged for creating new and tailor-made products, accelerating processing and - most of all - creating emulators for the entire production of global forecasts that outperform traditional systems at the spatial resolution of the training data. In this context, the role of both numerical models and observations is changing from being equation to data driven. Analyses and reanalyses are becoming the new currency for training machine learning, and operational centres are in a powerful position as they generate these datasets based on decades worth of experience. This environment creates incredible opportunities to progress much faster than in the past but also uncertainties about what the strategic implications on defining cost-effective and sustainable research and operations are, and how to achieve sufficient high-performance computing and data handling capacities. It will take individual national public services a while to understand what to focus on and how to coordinate their substantial investments in staff and infrastructure at institutional, national and international level. This paper addresses this new situation operational weather prediction finds itself in through formulating the most likely "what if?" scenarios for the near future and provides an outline for how weather centres could adapt.

physics.ao-ph↗

arXiv2023

Weakly Supervised Multi-Modal 3D Human Body Pose Estimation for Autonomous Driving

Accurate 3D human pose estimation (3D HPE) is crucial for enabling autonomous vehicles (AVs) to make informed decisions and respond proactively in critical road scenarios. Promising results of 3D HPE have been gained in several domains such as human-computer interaction, robotics, sports and medical analytics, often based on data collected in well-controlled laboratory environments. Nevertheless, the transfer of 3D HPE methods to AVs has received limited research attention, due to the challenges posed by obtaining accurate 3D pose annotations and the limited suitability of data from other domains. We present a simple yet efficient weakly supervised approach for 3D HPE in the AV context by employing a high-level sensor fusion between camera and LiDAR data. The weakly supervised setting enables training on the target datasets without any 2D/3D keypoint labels by using an off-the-shelf 2D joint extractor and pseudo labels generated from LiDAR to image projections. Our approach outperforms state-of-the-art results by up to $\sim$ 13% on the Waymo Open Dataset in the weakly supervised setting and achieves state-of-the-art results in the supervised setting.

cs.CV↗

arXiv2023

A Visual Quality Assessment Method for Raster Images in Scanned Document

Image quality assessment (IQA) is an active research area in the field of image processing. Most prior works focus on visual quality of natural images captured by cameras. In this paper, we explore visual quality of scanned documents, focusing on raster image areas. Different from many existing works which aim to estimate a visual quality score, we propose a machine learning based classification method to determine whether the visual quality of a scanned raster image at a given resolution setting is acceptable. We conduct a psychophysical study to determine the acceptability at different image resolutions based on human subject ratings and use them as the ground truth to train our machine learning model. However, this dataset is unbalanced as most images were rated as visually acceptable. To address the data imbalance problem, we introduce several noise models to simulate the degradation of image quality during the scanning process. Our results show that by including augmented data in training, we can significantly improve the performance of the classifier to determine whether the visual quality of raster images in a scanned document is acceptable or not for a given resolution setting.

eess.IV↗

arXiv2020

Impact of the experimental approach on the observed electronic energy loss for light keV ions in thin self-supporting films

Energy spectra of backscattered and transmitted ions with primary energies of 50 keV and 100 keV interacting with self-supporting foils were recorded with a Time-of-Flight Medium-Energy Ion Scattering setup in a single experiment. Self-supporting Au and W foils without backing material were used. For He ions transmitted through Au the spectrum of detected particles shows two distinct components corresponding to different energy losses in the film, whereas for protons no such phenomenon was observed. To determine the origin of these different contributions, measurements for different angles of incidence and scattering angles have been evaluated. The results suggest that the two components in the spectrum of transmitted He ions could be attributed to impact parameter dependent energy loss, being more prominent for He ions than for protons. The main origin of the necessary impact parameter selection along the different ion trajectories is expected to be texture in the Au-foils.

cond-mat.mtrl-sci↗

arXiv2019

Electronic excitation of transition metal nitrides by light ions with keV energies

We investigated the specific electronic energy deposition by protons and He ions with keV energies in different transition metal nitrides of technological interest. Data were obtained from two different time-of-flight ion scattering setups and show excellent agreement. For protons interacting with light nitrides, i.e. TiN, VN and CrN, very similar stopping cross sections per atom were found, which coincide with literature data of N2 gas for primary energies <= 25 keV. In case of the chemically rather similar nitrides with metal constituents from the 5th and 6th period, i.e. ZrN and HfN, the electronic stopping cross sections were measured to exceed what has been observed for molecular N2 gas. For He ions, electronic energy loss in all nitrides was found to be significantly higher compared to the equivalent data of N2 gas. Additionally, deviations from velocity proportionality of the observed specific electronic energy loss are observed. A comparison with predictions from density functional theory for protons and He ions yields a high apparent efficiency of electronic excitations of the target for the latter projectile. These findings are considered to indicate the contributions of additional mechanisms besides electron hole pair excitations, such as electron capture and loss processes of the projectile or promotion of target electrons in atomic collisions.

cond-mat.mtrl-sci↗

arXiv2019

Electronic interaction of slow hydrogen and helium ions in the nickel-silicon system

Electronic stopping cross sections (SCS) of nickel, silicon and nickel-silicon alloys for protons and helium (He) ions are studied in the regime of medium and low energy ion scattering, i.e., for ion energies in the range from 500 eV to 200 keV. For protons, at velocities below the Bohr velocity the deduced SCS is proportional to the ion velocity for all investigated materials. In contrast, for He ions non-linear velocity scaling is observed in all investigated materials. Static calculations using density functional theory (DFT) available from literature accurately predict the SCS of Ni and Ni-Si alloy in the regime with observed velocity proportionality. At higher energies, the energy dependence of the deduced SCS of Ni for protons and He ions agrees with the prediction by recent time dependent DFT calculations. The measured SCS of the Ni-Si alloy was compared to the SCS obtained from Bragg's rule based on SCS for Ni and Si deduced in this study, yielding good agreement for protons, but systematic deviations for He projectiles, by almost 20%. Overall, the obtained data indicate the importance of non adiabatic processes such as charge exchange for proper modelling of electronic stopping of in particular medium energy ions heavier than protons in solids.

physics.atom-ph↗

arXiv2019

On the influence of uncertainties in scattering potentials on quantitative analysis using keV ions

Experimental spectra from medium energy ion scattering were compared to Monte-Carlo simulations (employing the TRBS code) to obtain information on the scattering potential. The impact of uncertainties in the interatomic potential on quantification of sample properties such as thickness, composition or electronic stopping was investigated for different scattering geometries: backscattering and transmission. For backscattered He ions with tens of keV primary energy the scattering potential was found to overestimate the multiple scattering background in the energy spectra resulting in an uncertainty of < 3 % in quantitative analysis. Light ions transmitted through a sample for equivalent path length in the medium are only affected minorly by changes in the scattering potential. This effect becomes more distinct for heavier primary ions.

cond-mat.mtrl-sci↗

arXiv2019

Neutralization of slow helium ions scattered from single crystalline aluminum and tantalum surfaces and their oxides

We investigated the impact of surface oxygen on the ion yield for He$^+$ ions scattered from different single crystalline surfaces in low-energy ion scattering. Initially clean Al(111) and Ta(111) were exposed to molecular oxygen and ion spectra for different oxidation stages and different primary energies were recorded. A comparison of ion yields normalized to the differential scattering cross section as well as experimental factors allows obtaining information about the influence of oxygen on charge exchange processes. The decrease in the ion yield of both metals with exposure cannot be explained by different surface coverages exclusively, but requires the neutralization efficiency to be dependent on the chemical structure of the surface. For Ta, additionally, a different energy dependency of the ion yield obtained in the metal and oxide occurs. The ion yield for O shows in both surfaces a significantly weaker energy dependency than the investigated metals.

cond-mat.mtrl-sci↗

arXiv2017

Free-running Sn precipitates: an efficient phase separation mechanism for metastable GeSn epilayers

We report on the temperature stability of pseudomorphic GeSn films grown by molecular beam epitaxy on Ge(001) substrates. Both the growth temperature-dependence and the influence of post-growth annealing steps were investigated. In either case we observe that decomposition of metastable epilayers with Sn concentrations around 10% sets in above 230{\deg}C, the eutectic temperature of the Ge/Sn system. Time-resolved annealing experiments in a scanning electron microscope reveal the crucial role of liquid Sn droplets in this phase separation process. Driven by a gradient of the chemical potential, the Sn droplets move on the surface along preferential crystallographic directions, thereby taking up Sn and Ge from the strained GeSn layer at their leading edge. While Sn-uptake increases the volume of the melt, dissolved Ge becomes re-deposited by a liquid-phase epitaxial process at the trailing edge of the droplet. Secondary droplets are launched from the rims of the single-crystalline Ge trails into intact regions of the GeSn film, leading to an avalanche-like transformation front between the GeSn film and re-deposited Ge. This process makes phase separation of metastable GeSn layers particularly efficient at rather low temperatures.

cond-mat.mtrl-sci↗

arXiv2013

The potential of Rutherford Backscattering Spectrometry for composition analysis of colloidal nanocrystals

We investigate the potential of elastic scattering of energetic ions for compositional analysis of magnetic colloidal nanocrystals. Thin layers of nanocrystals deposited by spin coating on Si-wafers are investigated by two different set-ups for Rutherford Backscattering Spectrometry (RBS), employing different projectile ions (4He, 12C) and primary energies (600 keV - 8 MeV). The advantages and disadvantages of the different approaches are discussed in terms of obtainable mass resolution, necessary primary particle fluence and deposited energy. It is shown that different isotopes of transition metals can be resolved by employing 8 MeV 12C3+ primary ions.

cond-mat.mtrl-sci↗

arXiv2013

Electronic energy loss processes for slow H and He ions in metals and insulators: new insights

Electronic stopping of H and He ions in metals and insulators is analyzed at velocities below 0.2 atomic units, i.e. below 1 keV for H and below 4 keV for He. In metals, stopping of H ions is affected by d-electrons only when the d-band extends up to the Fermi energy; for He ions, also d-bands well below the Fermi energy contribute significantly to electronic stopping. In insulators, the low threshold velocity for electronic stopping cannot be explained by electron-hole pair excitation; charge exchange cycles, however, may govern the threshold behavior of electronic stopping in ionic crystals.

cond-mat.mtrl-sci↗