Offenburg University of Applied Sciences
Hochschulschriftenserver der Hochschule OffenburgNot a member yet
6641 research outputs found
Sort by
Digitalisierung von Raster-Zeichnungen mit Deep Learning // Digitization of Raster Drawings with Deep Learning
No full textDie automatische Extraktion von Produkt- und Fertigungsinformationen (Product Manufacturing Information - PMI) aus technischen CAD-Zeichnungen ist eine Voraussetzung für die Fertigung und Qualitätskontrolle in der Produktion. Aufgrund des speziellen Stils von CAD-Zeichnungen und der begrenzten Verfügbarkeit von Trainings- und Testdaten bleibt die Digitalisierung von CAD-Zeichnungen in Rasterbildern eine Herausforderung für Texterkennungssoftware (Optical Character Recognition - OCR). Dieser Beitrag stellt ein neuartiges, auf Deep Learning basierendes Framework vor, das dieses Problem adressiert, indem es Form- und Lagetoleranzen (Geometrical Dimensioning and Tolerancing - GD&T) sowie Bemaßungen in CAD-Zeichnungen lokalisiert und erkennt. Das Framework besteht aus einem zentralen Lokalisierungsmodul und mehreren nachgelagerten Pipelines für einzelne Klassen von PMI. Die Leistungsfähigkeit des Lokalisierungsmoduls, des Netzwerks zur Zeilenerkennung und der einzelnen Pipelines wird anhand realer Datensätze evaluiert. Ihre Leistung wird mit der des OCR-Programms Tesseract verglichen.Automatic extraction of Product Manufacturing Information (PMI) from mechanical CAD drawings is a prerequisite for manufacturing and production quality control. Because of the special style of CAD drawings and the limited availability of training and test data, digitizing CAD drawings in raster images remains a challenge for Optical Character Recognition (OCR) systems. This work presents a novel deep learning-based framework to address this problem, which localizes and recognizes Geometrical Dimensioning and Tolerancing (GD&T) and dimensions in CAD drawings. The framework is composed of a centralized localization module and several subsequent pipelines for individual classes of PMI. The performance of the localization module, the text recognition network and the individual pipelines is evaluated using real data sets. Their performance is compared with the performance of the OCR software Tesseract
A data-driven approach towards an e-health platform for dementia intervention
No full textGlobally, only 20-50% of dementia cases are diagnosed, indicating a major gap in the identification and treatment of this neurological disorder. Despite this, there is widespread agreement among both doctors and patients on the value of early diagnosis to adjust care and daily routines effectively. To address this issue, the German federal government launched a national dementia strategy in 2020. Current research, suggests that early treatment of certain risk factors could prevent or slow down one-third of dementia cases. The METIS project is based on the understanding that dementia is caused by complex processes which are linked to individual risk profiles and aims to make personalized treatment strategies for neurodegenerative diseases more accessible through an e-health platform
Die Auswirkung statischer interauraler Differenzen von Zeit oder Ort auf die Schalllokalisation bei bimodal Versorgten und Normalhörenden
No full textFrühere Arbeiten haben gezeigt, dass die Fähigkeit zur Lokalisation von Schallsignalen in der Horizontalebene bei Trägern von Cochlea-Implantat (CI) ipsi- und Hörgerät (HG) kontralateral (bimodale Versorgung), verbessert werden kann, wenn die statische interaurale Zeitverzögerung (Timing Mismatch – TM) technisch reduziert wird [1]. Das TM wird durch Unterschiede der Prozessierungslatenzen von CI und HG und durch unterschiedliche Stimulationsorte (CI: Hörnerv; HG: äußerer Gehörgang) verursacht. Ein CI-Hersteller hat die technische Reduktion des TM zwischenzeitlich als frequenzübergreifende Verzögerung der CI-Stimulation für die CI-Anpassung verfügbar gemacht. Dadurch lässt sich das TM zwar reduzieren, für eine Eliminierung bräuchte es aber eine frequenzspezifische Verzögerung der CI-Stimulation.
Zusätzlich zum TM treten bei bimodal Versorgten auch statische interaurale tonotope Differenzen auf (Place of Stimulation Mismatch – PM). Grund hierfür ist die begrenzte Insertionstiefe des CI-Elektrodenarrays. Meist wird diesem Umstand bei der CI-Anpassung nicht Rechnung getragen. Es wird eine von der Insertionstiefe unabhängige Frequenz-Elektroden-Zuordnung verwendet. Da in einem akustisch hörenden Ohr aber jedem Ort in der Cochlea nicht nur eine charakteristische Frequenz sondern auch eine charakteristische Prozessierungszeit (Cochlear Delay) zugeordnet werden kann, gelingt die Feinabstimmung von TM und PM bei bimodaler Versorgung nur gemeinsam. Als erster Ansatz zur Untersuchung dieses Effektes wird im Vortrag die Auswirkung von TM und PM separat vorgestellt. Erhöhung des TM wie auch des PM führt zur Reduktion der Schalllokalisationsfähigkeit. Allerdings reduziert TM sowohl den RMSE als auch den Signed Bias, während mit steigendem PM primär der RMSE schlechter wird. Wie TM/PM und binaurale Fusion interagiert, wird zusätzlich im Vortrag erörtert
Angle of Arrival versus Multipath Propagation Method for Curbstone Height Estimation in Automotive Radar Application
No full textTarget height estimation plays an important role in automotive radar parking scenarios or autonomous driving. For example, knowing a curb's height can assist in deciding whether it is traversable as well as avoiding collision. In this paper, two signal processing methods for this purpose are assessed. The first approach calculates the angle of arrival (AoA) of the received signals by processing the virtual channels offered by multiple-input multiple-output (MIMO) antennas. In the second method, the radar wave multipath propagation model in combination with super-resolution algorithms, namely ESPRIT and MUSIC, is exploited to calculate the curb height considering the geometry of antennas. The theory behind this method is discussed in detail, and the results are evaluated using a 77 GHz Frequency Modulated Continuous Wave (FMCW) radar
Decarburization and Gas Formation During Sintering of Alloyed PM Steel Components
Get PDFThis study investigates the delubrication, reduction, and decarburization processes of powder metallurgical steel alloys (CrM, CrL, AHC, Mo85, SintD 35) and an unalloyed steel during sintering in a pure hydrogen atmosphere. Utilizing in-situ FTIR gas phase analysis, components with ethylenebisstearamide (EBS) as a lubricant are analyzed. EBS decomposition in steel components yields CO, CO2, H2O, and CH4, with dominant CH groups observed in the 230 °C to 480 °C range. In the temperature range between 750 °C and 850 °C, where CO formation is expected due to the reduction of surface iron oxides, CH4 is present instead, indicating that an “internal getter effect” also occurs in pre-alloyed powders. In addition, with high carbon activity, the reduction of internal iron oxides and the reduction of chromium oxides also trigger an internal getter effect. Depending on the carbon potential, these processes cause a considerable reduction in the carbon content of the powder metallurgical components. The study therefore shows that the decarburization of powder metallurgical components during the heat treatment phases prior to sintering in a 100 pct hydrogen atmosphere is less due to the mechanism of delubrication, but rather to mechanisms of carbothermal reduction
The Influence of Hyperparameters on GANs Performance for Medical Image Transformation
No full textGenerative Adversarial Networks (GANs) have accomplished compelling performance in many fields such as image-to-image transformation, image data generation, translation of image information into text information, and many more. In GANs, two neural networks, a generator and discriminator, compete with each other in an adversarial manner, and due to their robust generating ability, GANs can produce high-quality images for human reference. Thus, GANs are a very promising approach artificial intelligence field. In this work, GANs with optimized hyperparameters selection are proposed to generate skin lesion medical images that are indistinguishable from the real images and analyze the quality of generated images using image quality metrics including Peak Signal-to-Noise Ratio (PSNR), Structural Similarity Index (SSIM), and Fréchet Inception Distance (FID). Additionally, the impact of hyperparameters including learning rate, batch size, latent space size, and number of epochs is analyzed against GANs performance. The experimental results indicate that the learning rate is the most effective hyperparameter for GAN stability and performance. Moreover, latent dimension, batch size, and number of epochs have a relatively small impact when paired with an appropriate learning rate thus good learning rate provides flexibility in optimizing these hyperparameters
Wirtschafts- und Steuerstrafrecht
No full textWirtschafts- und Steuerstrafrecht in einem Werk gebündelt.
Alles für die wirtschaftsrechtliche Praxis
Der Kommentar erläutert alle relevanten straf- und ordnungswidrigkeitsrechtlichen Normen, Gesetze, Verordnungen und Richtlinien zum Wirtschafts- und Steuerstrafrecht in einem Band und bietet einen verlässlichen Überblick über den aktuellen Meinungsstand. Fundiert und praxisgerecht dargestellt werden:
- die wirtschaftsstrafrechtlichen Normen des StGB
- Nebengesetze und Verordnungen mit strafrechtlichen Normen (z.B. Lebensmittelstrafrecht)
- die strafrechtlichen Normen im Steuerrecht
- die gesetzlichen Änderungen der Regelungen über die strafbefreiende - Selbstanzeige einschließlich der Regelung über die Einstellung des Verfahrens nach Selbstanzeige (§ 398a AO)
- europarechtliche Normen und Richtlinien und ihre Auswirkungen auf das deutsche Recht
- gesellschafts-, wertpapier-, urheber- und wettbewerbsrechtliche Straf-, Ordnungswidrigkeits- und Steuervorschriften
Zur Neuauflage
Der Kommentar erscheint erstmalig in zwei Bänden und ermöglicht dadurch noch detailliertere und umfassendere Erläuterungen. Zudem wird der bewährte Kommentar um zahlreiche Aspekte erweitert, u.a. mit der Kommentierung des § 24 LkSG, § 23 GeschGehG, § 21 MiLoG,§§ 42, 43 DS-GVO, § 42 StaRUG und der erweiterten Kommentierung des GWB
Are Vision Language Models Texture or Shape Biased and Can We Steer Them?
No full textVision language models (VLMs) have drastically changed the computer vision model landscape in only a few years, opening an exciting array of new applications from zero-shot image classification, over to image captioning, and visual question answering. Unlike pure vision models, they offer an intuitive way to access visual content through language prompting. The wide applicability of such models encourages us to ask whether they also align with human vision - specifically, how far they adopt human-induced visual biases through multimodal fusion, or whether they simply inherit biases from pure vision models. One important visual bias is the texture vs. shape bias, or the dominance of local over global information. In this paper, we study this bias in a wide range of popular VLMs. Interestingly, we find that VLMs are often more shape-biased than their vision encoders, indicating that visual biases are modulated to some extent through text in multimodal models. If text does indeed influence visual biases, this suggests that we may be able to steer visual biases not just through visual input but also through language: a hypothesis that we confirm through extensive experiments. For instance, we are able to steer shape bias from as low as 49% to as high as 72% through prompting alone. For now, the strong human bias towards shape (96%) remains out of reach for all tested VLMs
Decoding dynamic interactions between EGFR-TKD and DAC through computational and experimental approaches
Get PDFIn the quest for effective lung cancer treatments, the potential of 3,6-diaminoacridine-9-carbonitrile (DAC) has emerged as a game changer. While DAC's efficacy against glioblastoma is well documented, its role in combating lung cancer has remained largely untapped. This study focuses on CTX-1, exploring its interaction with the pivotal EGFR-TKD protein, a crucial target in lung cancer therapeutics. A meticulous molecular docking analysis revealed that CTX-1 exhibits a noteworthy binding affinity of −7.9 kcal/mol, challenging Erlotinib, a conventional lung cancer medication, which displayed a binding affinity of −7.3 kcal/mol. For a deeper understanding of CTX-1's molecular mechanics, this study employed rigorous 100-ns molecular dynamics simulations, demonstrating CTX-1's remarkable stability in comparison with erlotinib. The Molecular Mechanics Poisson–Boltzmann Surface Area (MM-PBSA) method further corroborated these results, with CTX-1 showing a free binding energy of −105.976 ± 1.916 kJ/mol. The true prowess of CTX-1 was tested against diverse lung cancer cell lines, including A549, Hop-62 and H-1299. CTX-1 not only significantly outperformed erlotinib in anticancer activity but also exhibited a spectrum of therapeutic effects. It effectively diminished cancer cell viability, induced DNA damage, halted cell cycle progression, generated reactive oxygen species (ROS), impaired mitochondrial transmembrane potential, instigated apoptosis and successfully inhibited EGFR-TKD. This study not only underscores the potential of CTX-1 a formidable contender in lung cancer treatment but also marks a paradigm shift in oncological therapeutics, offering new horizons in the fight against this formidable disease