A5101764191- Parallel Computing and Optimization Techniques
- Advanced Data Storage Technologies
- Distributed and Parallel Computing Systems
- Advanced Neural Network Applications
- Adversarial Robustness in Machine Learning
- Video Surveillance and Tracking Methods
- Domain Adaptation and Few-Shot Learning
- Algorithms and Data Compression
- Electromagnetic Scattering and Analysis
- Radiomics and Machine Learning in Medical Imaging
- Advanced Vision and Imaging
- Scientific Research and Discoveries
- Simulation Techniques and Applications
- Human Pose and Action Recognition
- Cloud Computing and Resource Management
- Image Enhancement Techniques
- AI in cancer detection
- Scientific Computing and Data Management
- Medical Image Segmentation Techniques
- Theoretical and Computational Physics
- Computer Graphics and Visualization Techniques
- Advanced Malware Detection Techniques
- Anomaly Detection Techniques and Applications
- Particle accelerators and beam dynamics
- Neural Networks and Applications
TomTom (Netherlands)
2018-2021
Robert Bosch (Germany)
2017-2020
Chemnitz University of Technology
2009-2018
Weierstrass Institute for Applied Analysis and Stochastics
2014-2016
Graz University of Technology
2007-2016
Technische Universität Berlin
2014
University of Amsterdam
2011
Ministry of Defence
2009
Humboldt-Universität zu Berlin
2009
Ludwig-Maximilians-Universität München
2006
Based on a parallel scalable library for Coulomb interactions in particle systems, comparison between the fast multipole method (FMM), multigrid-based methods, Fourier transform (FFT)-based and Maxwell solver is provided case of three-dimensional periodic boundary conditions. These methods are directly compared with respect to complexity, scalability, performance, accuracy. To ensure comparable conditions all cover typical applications, we tested same set computers using identical benchmark...
We introduce a framework for unconstrained 3D human upper body pose estimation from multiple camera views in complex environment. Its main novelty lies the integration of three components: single-frame recovery, temporal and model texture adaptation. Single-frame recovery consists hypothesis generation stage, which candidate poses are generated, based on probabilistic hierarchical shape matching each view. In subsequent verification re-projected into other ranked according to multi-view...
Traffic light detection is crucial for environment perception and decision-making in autonomous driving. State-of-the-art detectors are built upon deep Convolutional Neural Networks (CNNs) have exhibited promising performance. However, one looming concern with CNN based how to thoroughly evaluate the performance of accuracy robustness before they can be deployed vehicles. In this work, we propose a visual analytics system, VATLD, equipped disentangled representation learning semantic...
Based on the Langevin description of Continuous Time Random Walk (CTRW), we consider a generalization CTRW in which waiting times between subsequent jumps are correlated. We discuss cases exponential and slowly decaying persistent power-law correlations as two generic examples obtain corresponding mean squared displacements functions time. In case exponential-type (sub)diffusion at short is slower than absence correlations. At long behavior displacement same uncorrelated CTRW. For find...
Increasing the efficiency of production and manufacturing processes is a key goal initiatives like Industry 4.0. Within context European research project ARROWHEAD, we enable secure smart maintenance services. An overall to proactively predict optimize Maintenance, Repair Operations (MRO) carried out by device maintainer, for industrial devices deployed at customer. Therefore it necessary centrally acquire relevant equipment status data from remotely located over Internet. Consequently,...
We present a system for the estimation of unconstrained 3D human upper body movement from multiple cameras. Its main novelty lies in integration three components: single frame pose recovery, temporal and model adaptation. Single recovery consists hypothesis generation stage, where candidate poses are generated based on hierarchical shape matching individual camera views. In subsequent verification reprojected to other views ranked according multiview score. Temporal computing best...
Software implementations of modern block ciphers often require large lookup tables along with code size increasing optimizations like loop unrolling to reach peak performance on general-purpose processors. Therefore, are difficult implement efficiently embedded devices cell phones or sensor nodes where run-time memory and program ROM scarce resources. In this paper, the performance, energy consumption, runtime requirements, five RC6, Rijndael, Serpent, Twofish, XTEA StrongARM SA-1100...
Abstract Percutaneous radiofrequency ablation (RFA) is a minimally invasive technique that destroys cancer cells by heat. The heat results from focusing energy in the spectrum through needle. Amongst others, this can enable treatment of patients who are not eligible for an open surgery. However, possibility recurrent liver due to incomplete tumor makes post-interventional monitoring via regular follow-up scans mandatory. These have be carefully inspected any conspicuousness. Within study, RF...
Semantic segmentation is a task that traditionally requires large dataset of pixel-level ground truth labels, which time-consuming and expensive to obtain. Recent advancements in the weakly-supervised setting show reasonable performance can be obtained by using only image-level labels. Classification often used as proxy train deep neural network from attention maps are extracted. However, classification needs minimum evidence make predictions, hence it focuses on most discriminative object...
The power consumption of programs and algorithms is currently a very active research field. This includes the investigation effect different programming techniques on consumption. Some have already been studied intensively. However, there are that did not get as much attention needed so far. One these vectorization programs, which uses special operations to calculate several data in one step. In this article, we investigate study program versions dense matrix multiplication combine with...
We apply a flexible numerical integrator to the simulation of adiabatic quantum computation with nonlinear paths. find that path may significantly improve performance algorithms versus conventional straight-line interpolations. The employed is suitable for solving time-dependent Schr\"odinger equation any qubit Hamiltonian. Its storage format reduces cost and matrix-vector multiplication in comparison common sparse matrix schemes.
Ultrasound (US) is the most commonly used liver imaging modality worldwide. It plays an important role in follow-up of cancer patients with metastases. We present interactive segmentation approach for tumors US acquisitions. Due to low image quality and contrast between surrounding tissue images, very challenging. Thus, clinical practice still relies on manual measurement outlining images. target this problem by applying algorithm data, allowing user get real-time feedback results. The has...
In this contribution, we present a semi-automatic segmentation algorithm for radiofrequency ablation (RFA) zones via optimal s-t-cuts. Our interactive graph-based approach builds upon polyhedron to construct the graph and was specifically designed computed tomography (CT) acquisitions from patients that had RFA treatments of Hepatocellular Carcinomas (HCC). For evaluation, used twelve post-interventional CT datasets clinical routine as evaluation metric utilized Dice Similarity Coefficient...
The Fast Multipole Method (FMM) is an efficient, widely used method for the solution of N-body problems. One main data structures a hierarchical tree structure describing separation into near-field and far-field particle interactions. This article presents automatic tuning FMM by selecting optimal depth based on integrated performance prediction computations. exploits benchmarking significant parts implementation to adapt specific hardware system being used. Furthermore, separate analysis...
Adversarial training has been recently employed for realizing structured semantic segmentation, in which the aim is to preserve higher-level scene structural consistencies dense predictions. However, as we show, value-based discrimination between predictions from segmentation network and ground-truth annotations can hinder process learning improve qualities well disabling properly expressing uncertainties. In this paper, rethink adversarial propose reformulate fake/real framework with a...
Traffic light detection is crucial for environment perception and decision-making in autonomous driving. State-of-the-art detectors are built upon deep Convolutional Neural Networks (CNNs) have exhibited promising performance. However, one looming concern with CNN based how to thoroughly evaluate the performance of accuracy robustness before they can be deployed vehicles. In this work, we propose a visual analytics system, VATLD, equipped disentangled representation learning semantic...
Parallel sorting methods for distributed memory systems often use partitioning algorithms to prepare the redistribution of data items. This article proposes a algorithm that calculates specified by number items be finally located on each process. can also used with weights, which might express computational load expected, and produce an individual accumulated weight Another important feature is sets duplicated keys handled. those properties needed parallel scientific application codes, such...
This article investigates two particle data redistribution methods for the coupling of application-independent solvers long range interactions with a dynamics simulation. The rely on their own processing techniques and domain decomposition schemes, but are implemented within single parallel software library unique interface. Thus, efficient reordering required transfer between an application interface as well solver. first method hides all inside restores original order distribution. second...