检索结果-内蒙古大学图书馆

The Norm Must Go On: Dynamic Unsupervised Domain Adaptation by Normalization

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Mirza, Muhammad Jehanzeb Micorek, Jakub Possegger, Horst Bischof, Horst Institute for Computer Graphics and Vision Graz University of Technology Austria Christian Doppler Laboratory for Embedded Machine Learning Austria

Domain adaptation is crucial to adapt a learned model to new scenarios, such as domain shifts or changing data distributions. Current approaches usually require a large amount of labeled or unlabeled data from the shifted domain. This can be a hurdle in fields which require continuous dynamic adaptation or suffer from scarcity of data, e.g. autonomous driving in challenging weather conditions. To address this problem of continuous adaptation to distribution shifts, we propose Dynamic Unsupervised Adaptation (DUA). By continuously adapting the statistics of the batch normalization layers we modify the feature representations of the model. We show that by sequentially adapting a model with only a fraction of unlabeled data, a strong performance gain can be achieved. With even less than 1% of unlabeled data from the target domain, DUA already achieves competitive results to strong baselines. In addition, the computational overhead is minimal in contrast to previous approaches. Our approach is simple, yet effective and can be applied to any architecture which uses batch normalization. We show the utility of DUA by evaluating it on a variety of domain adaptation datasets and tasks including object recognition, digit recognition and object detection. © 2021, CC BY.

关键词： Object recognition

HO-3D v3: Improving the accuracy of hand-object annotations of the HO-3D dataset

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Hampali, Shreyas Sarkar, Sayan Deb Lepetit, Vincent Institute for Computer Graphics and Vision Graz University of Technology Graz Austria Université Paris-Est École des Ponts ParisTech Paris France

A Queueing-theoretic Analysis of the Performance of a Cloud Computing Infrastructure: Accounting for Task Reneging or Dropping

学校读者我要写书评

暂无评论

A Queueing-theoretic Analysis of the Performance of a Cloud ...

2022 International Conference on Electrical, computer, Communications and Mechatronics Engineering, ICECCME 2022

作者： Kuaban, Godlove Suila Soodan, Bhavneet Singh Kumar, Rakesh Czekalski, Piotr Institute of Theoretical and Applied Informatics Polish Academy of Sciences Baltycka 5 Gliwice44-100 Poland Chandigarh University Department of Mathematics Punjab 140413 India Namibia University of Science and Technology Department of Mathematics and Statistics Namibia Department of Computer Graphics Vision and Digital Systems Silesian University of Technology Akademicka 16 Gliwice44-100 Poland

ISBN: (数字)9781665470957

ISBN: (纸本)9781665470957

Cloud computing has revolutionized the information technology era. It offers high-speed computing, storage, and ICT resources on demand. One of the significant challenges in cloud computing is the impact of impatient users or request (task) reneging. When a request has been compromised, missed its execution deadline, or depends on other rejected ones, it must be removed from the queue without being processed. The reneging or removal of tasks from queues may trigger the reneging or removal of other tasks that depend of them. We refer to this reneging of requests from the load balancing or computing queues as correlated reneging. We presented the performance analysis of a network of queues that constitute a queueing model of a simplified cloud computing infrastructure. We show the relationship between the load, delay, and probability of buffer saturation (blocking probability). It is seen that at about 80% utilization, a small increase in utilization (due to a small increase in the arrival rate Δλ for a fixed service rate μ) results in a sharp increase in the delay experienced by the tasks, and on the probability of task rejection or blocking. © 2022 IEEE.

关键词： Cloud computing

Adaptation of deep learning models to distinguish species with strongly overlapping characteristics: the example of spores of the genus Riccia

学校读者我要写书评

暂无评论

Ecological Informatics 2025年 90卷

作者： Herdy, Stefan Pöltl, Martina Berg, Christian Weber, Bettina Department of Biology Holteigasse 6 University of Graz Graz Austria Institute of Computer Graphics and Vision Graz University of Technology Graz Austria Studienzentrum Naturkunde Universalmuseum Joanneum Graz 8045 Austria

The liverwort genus Riccia is fairly characteristic in itself, but differentiation between some of the species herein proved as complex. Morphological thallus characteristics show a wide overlap and variation, also influenced by environmental conditions, whereas spore characteristics might be more consistent within individual species. Thus, here we investigated if morphological spore characteristics can be analyzed by means of generative as well as discriminative deep learning models, allowing a differentiation between very similar species. We applied a modified Generative Adversarial Network on spore images of the genus Riccia to generate images, that have a better expression of class specific morphological features. We also created one single spore image for every species that comprises the representative species characteristics. This approach allowed us to also distinguish morphologically very similar taxa and to quantify the morphological similarity between them. Our results show that generative modeling can improve morphological species classification in biology and provides a framework to quantify qualitative species features and similarities. These new tools facilitate more accurate and efficient species identification, thus greatly advancing methodologies in taxonomic research and environmental monitoring. © 2024

关键词： Deep learning GAN Liverworts Machine vision WGAN

Fuzzy Querying in the Cloud-based Environment for Data Stream-driven Predictive Maintenance in AGV-enabled Smart Factories

学校读者我要写书评

暂无评论

Fuzzy Querying in the Cloud-based Environment for Data Strea...

IEEE International Conference on Big Data

作者： Bożena Małysiak-Mrozek Dominik Romanów Piotr Grzesik Paweł Benecki Alexandre Niyomugaba Theodore Habimana Daniel Kostrzewa Krzysztof Tokarz Che-Lun Hung Dariusz Mrozek Department of Distributed Systems and Informatics Devices Silesian University of Technology Gliwice Poland Department of Applied Informatics Silesian University of Technology Gliwice Poland Department of Computer Science INES Ruhengeri - Institute of Applied Sciences Rwanda Department of Graphics Computer Vision and Digital Systems Silesian University of Technology Gliwice Poland Institute of Biomedical Informatics National Yang Ming Chiao Tung University Taipei Taiwan

ISBN: (数字)9798350362480

ISBN: (纸本)9798350362497

Fuzzy data processing enables data enrichment and increases data interpretation in industrial environments. In the cloud-based IoT data ingestion pipelines, fuzzy data processing can be implemented in several locations, closer to the IoT events gateways, stream processors, or the persistence layer before the data is visualized. Since Automated Guided Vehicles (AGV)-enabled manufacturing can produce vast amounts of data, the decision on the placement of the fuzzy data processing can be important for secondary processes performed on the enriched data, like the predictive maintenance inferencing. In this paper, we analyze two locations of fuzzy data processing in the cloud-based environment built for monitoring AGVs in smart factories - by formulating fuzzy queries against data streams on stream processing units and data at rest in a database. The querying scenarios cover fuzzy filtering with simple and complex criteria, fuzzy filtering through assignment to a linguistic variable, and joining data streams by representing joining attributes as fuzzy numbers. The experimental results show that querying the data stream can be more efficient and profitable in the scalable environment of many AGVs. However, the enrichment provided for the data at rest is also beneficial when gathering data for building future predictive maintenance models.

关键词： Remotely guided vehicles Program processors Filtering Pipelines Predictive models Logic gates Data processing Monitoring Predictive maintenance Smart manufacturing

Monte carlo scene search for 3D scene understanding

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Hampali, Shreyas Stekovic, Sinisa Sarkar, Sayan Deb Kumar, Chetan S. Fraundorfer, Friedrich Lepetit, Vincent Institute for Computer Graphics and Vision Graz University of Technology Graz Austria Université Paris-Est École des Ponts ParisTech Paris France

We explore how a general AI algorithm can be used for 3D scene understanding to reduce the need for training data. More exactly, we propose a modification of the Monte Carlo Tree Search (MCTS) algorithm to retrieve objects and room layouts from noisy RGB-D scans. While MCTS was developed as a game-playing algorithm, we show it can also be used for complex perception problems. Our adapted MCTS algorithm has few easy-to-tune hyperparameters and can optimise general losses. We use it to optimise the posterior probability of objects and room layout hypotheses given the RGB-D data. This results in an analysis-by-synthesis approach that explores the solution space by rendering the current solution and comparing it to the RGB-D observations. To perform this exploration even more efficiently, we propose simple changes to the standard MCTS' tree construction and exploration policy. We demonstrate our approach on the ScanNet dataset. Our method often retrieves configurations that are better than some manual annotations, especially on layouts. Copyright © 2021, The Authors. All rights reserved.

关键词： Trees (mathematics)

Deep Reinforcement Learning for Localization of the Aortic Annulus in Patients with Aortic Dissection 2nd

学校读者我要写书评

暂无评论

Deep Reinforcement Learning for Localization of the Aortic A...

2nd International Workshop on Thoracic Image Analysis, TIA 2020 Held in Conjunction with Medical Image Computing and computer-Assisted Intervention Conference, MICCAI 2020

作者： Codari, Marina Pepe, Antonio Mistelbauer, Gabriel Mastrodicasa, Domenico Walters, Shannon Willemink, Martin J. Fleischmann, Dominik Department of Radiology Stanford University School of Medicine StanfordCA United States Institute of Computer Graphics and Vision Graz University of Technology Graz Austria Department of Simulation and Graphics Otto-von-Guericke University Magdeburg Germany

ISBN: (纸本)9783030624682

Accurate localization of the aortic annulus is key to several imaging tasks, like cross-sectional aortic valve plane estimation, aortic root segmentation, and annulus diameter measurements. In this project, we propose an end-to-end trainable deep reinforcement learning (DRL) algorithm aimed at identification of the aortic annulus in patients with aortic dissection. We trained 5 different agents on a dataset of 75 CT scans from 66 patients following a sequential model-upgrading strategy. We evaluated the effect of performing different image preprocessing steps, adding batch normalization and regularization layers, and changing terminal state definition. At each step of this sequential process, the model performance has been evaluated on a validation sample composed of 24 CTA scans from 24 independent patients. Localization accuracy was defined as the Euclidean distance between estimated and target aortic annulus locations. Best model results show a median localization error equal to 2.98 mm with an interquartile range equal to [2.25, 3.81] mm, and a failure rate (i.e., percentage of samples with localization error in validation data. We proved the feasibility of DRL application for aortic annulus localization in CTA images of patients with aortic dissection, which are characterized by a large variability in aortic morphology and image quality. Nevertheless, further improvements are needed to reach expert-human level performance. © 2020, Springer Nature Switzerland AG.

关键词： Reinforcement learning

MonteFloor: Extending MCTS for reconstructing accurate large-scale floor plans

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Stekovic, Sinisa Rad, Mahdi Fraundorfer, Friedrich Lepetit, Vincent Institute for Computer Graphics and Vision Graz University of Technology Graz Austria Université Paris-Est École des Ponts ParisTech Paris France

We propose a novel method for reconstructing floor plans from noisy 3D point clouds. Our main contribution is a principled approach that relies on the Monte Carlo Tree Search (MCTS) algorithm to maximize a suitable objective function efficiently despite the complexity of the problem. Like previous work, we first project the input point cloud to a top view to create a density map and extract room proposals from it. Our method selects and optimizes the polygonal shapes of these room proposals jointly to fit the density map and outputs an accurate vectorized floor map even for large complex scenes. To do this, we adapt MCTS, an algorithm originally designed to learn to play games, to select the room proposals by maximizing an objective function combining the fitness with the density map as predicted by a deep network and regularizing terms on the room shapes. We also introduce a refinement step to MCTS that adjusts the shape of the room proposals. For this step, we propose a novel differentiable method for rendering the polygonal shapes of these proposals. We evaluate our method on the recent and challenging Structured3D and Floor-SP datasets and show a significant improvement over the state-of-the-art, without imposing any hard constraints nor assumptions on the floor plan configurations. Copyright © 2021, The Authors. All rights reserved.

关键词： Floors

Semi-supervised Virtual Regression of Aortic Dissections Using 3D Generative Inpainting 2nd

学校读者我要写书评

暂无评论

Semi-supervised Virtual Regression of Aortic Dissections Usi...

2nd International Workshop on Thoracic Image Analysis, TIA 2020 Held in Conjunction with Medical Image Computing and computer-Assisted Intervention Conference, MICCAI 2020

作者： Pepe, Antonio Mistelbauer, Gabriel Gsaxner, Christina Li, Jianning Fleischmann, Dominik Schmalstieg, Dieter Egger, Jan Institute of Computer Graphics and Vision Graz University of Technology Graz Austria School of Medicine Department of Radiology Stanford University StanfordCA United States Computer Algorithms in Medicine Laboratory Graz Austria Department of Simulation and Graphics University of Magdeburg Magdeburg Germany Department of Oral and Maxillofacial Surgery Medical University of Graz Graz Austria

ISBN: (纸本)9783030624682

Aortic dissection (AD) is a condition of the main artery of the human body, resulting in the formation of a new flow channel, or false lumen (FL). The disease is usually diagnosed with a computed tomography angiography (CTA) scan during the acute phase. A better understanding of the causes of AD requires knowledge of aortic geometry prior to the event, which is available only in very rare circumstances. In this work, we propose an approach to reconstruct the aorta before the formation of a dissection by performing 3D inpainting with a two-stage generative adversarial network (GAN). In the first stage of our two-stage GAN, a network is trained on the 3D edge information of the healthy aorta to reconstruct the aortic wall. The second stage infers the image information of the aorta to reconstruct the entire dataset. We train our two-stage GAN with 3D patches from 55 non-dissected aortic datasets and evaluate it on 20 more non-dissected datasets, demonstrating that our proposed 3D architecture outperforms its 2D counterpart. To obtain pre-dissection aortae, we mask the entire FL in AD datasets. Finally, we provide qualitative feedback from a renown expert on the obtained pre-dissection cases. © 2020, Springer Nature Switzerland AG.

关键词： Generative adversarial networks