Offenburg University of Applied Sciences
Hochschulschriftenserver der Hochschule OffenburgNot a member yet
6641 research outputs found
Sort by
As large as it gets - Studying Infinitely Large Convolutions via Neural Implicit Frequency Filters [+ Supplementary Material]
No full textRecent work in neural networks for image classification has seen a strong tendency towards increasing the spatial context during encoding. Whether achieved through large convolution kernels or self-attention, models scale poorly with the increased spatial context, such that the improved model accuracy often comes at significant costs. In this paper, we propose a module for studying the effective filter size of convolutional neural networks (CNNs). To facilitate such a study, several challenges need to be addressed: (i) we need an effective means to train models with large filters (potentially as large as the input data) without increasing the number of learnable parameters, (ii) the employed convolution operation should be a plug-and-play module that can replace conventional convolutions in a CNN and allow for an efficient implementation in current frameworks, (iii) the study of filter sizes has to be decoupled from other aspects such as the network width or the number of learnable parameters, and (iv) the cost of the convolution operation itself has to remain manageable i.e. we can not naïvely increase the size of the convolution kernel. To address these challenges, we propose to learn the frequency representations of filter weights as neural implicit functions, such that the better scalability of the convolution in the frequency domain can be leveraged. Additionally, due to the implementation of the proposed neural implicit function, even large and expressive spatial filters can be parameterized by only a few learnable weights. Interestingly, our analysis shows that, although the proposed networks could learn very large convolution kernels, the learned filters are well localized and relatively small in practice when transformed from the frequency to the spatial domain. We anticipate that our analysis of individually optimized filter sizes will allow for more efficient, yet effective, models in the future
Automating Wood Species Detection and Classification in Microscopic Images of Fibrous Materials with Deep Learning
No full textWe have developed a methodology for the systematic generation of a large image dataset of macerated wood references, which we used to generate image data for nine hardwood genera. This is the basis for a substantial approach to automate, for the first time, the identification of hardwood species in microscopic images of fibrous materials by deep learning. Our methodology includes a flexible pipeline for easy annotation of vessel elements. We compare the performance of different neural network architectures and hyperparameters. Our proposed method performs similarly well to human experts. In the future, this will improve controls on global wood fiber product flows to protect forests
Biomechanical validation of novel polyurethane-resin synthetic osteoporotic femoral bones in axial compression, four-point bending and torsion
No full textIn addition to human donor bones, bone models made of synthetic materials are the gold standard substitutes for biomechanical testing of osteosyntheses. However, commercially available artificial bone models are not able to adequately reproduce the mechanical properties of human bone, especially not human osteoporotic bone. To overcome this issue, new types of polyurethane-based synthetic osteoporotic bone models have been developed. Its base materials for the cancellous bone portion and for the cortical portion have already been morphologically and mechanically validated against human bone. Thus, the aim of this study was to combine the two validated base materials for the two bone components to produce femur models with real human geometry, one with a hollow intramedullary canal and one with an intramedullary canal filled with synthetic cancellous bone, and mechanically validate them in comparison to fresh frozen human bone. These custom-made synthetic bone models were fabricated from a computer-tomography data set in a 2-step casting process to achieve not only the real geometry but also realistic cortical thicknesses of the femur. The synthetic bones were tested for axial compression, four-point bending in two planes, and torsion and validated against human osteoporotic bone. The results showed that the mechanical properties of the polyurethane-based synthetic bone models with hollow intramedullary canals are in the range of those of the human osteoporotic femur. Both, the femur models with the hollow and spongy-bone-filled intramedullary canal, showed no substantial differences in bending stiffness and axial compression stiffness compared to human osteoporotic bone. Torsional stiffnesses were slightly higher but within the range of human osteoporotic femurs. Concluding, this study shows that the innovative polyurethane-based femur models are comparable to human bones in terms of bending, axial compression, and torsional stiffness
Unleashing Modified Deep Learning Models in Efficient COVID-19 Detection
No full textCOVID-19 is a unique and devastating respiratory disease outbreak that has affected global populations as the disease spreads rapidly. Many deep learning breakthroughs may improve COVID-19 prediction and forecasting as a tool for precise and fast detection. In this study, the dataset used contained 8055 CT image samples, 5427 of which were COVID cases and 2628 non-COVID. Again, 9544 X-ray samples included 4044 COVID patients and 5500 non-COVID cases. MobileNetV3, DenseNet201, and GoogleNet InceptionV1 show the highest accuracy of 97.872%, 97.567%, and 97.643%, respectively. The high accuracy indicates that these models can make many accurate predictions, as well as others, are also high for MobileNetV3 and DenseNet201. An extensive evaluation using accuracy, precision, and recall allows a comprehensive comparison to improve predictive models by combining loss optimization with scalable batch normalization. This research shows that these tactics improve model performance and resilience for advancing COVID-19 prediction and detection and show how deep learning can improve disease handling. The methods suggested in this research would strengthen healthcare systems, policymakers, and researchers to make educated decisions to reduce COVID-19 and other contagious diseases
Social-Media-Marketing
No full textVor dem Hintergrund der wachsenden Bedeutung von Social Media für das Marketing von Unternehmen werden in diesem Beitrag zunächst einige wichtige Rahmenbedingungen und Entwicklungen für das Social-Media-Marketing skizziert. Da ein erfolgreiches Social-Media-Marketing immer auch ein strukturiertes und durchdachtes Vorgehen erfordert, werden ein diesbezüglicher Planungsprozess und seine Phasen vorgestellt. Daneben werden zentrale Implementierungsoptionen beschrieben. Hierzu gehören Werbung (über Plattformen und Influencer), Kundenservice, Community-Management, Social Recruitment, interne Nutzung und Business Profile
CSDP-2024 : Cyber Security and Data Protection 2024
No full textThere were 27 papers submitted for peer-review to this workshop. Out of these, 13 papers were accepted for this volume, 7 as regular papers and 6 as short papers
