检索结果-内蒙古大学图书馆

Exploiting autoencoder’s weakness to generate pseudo anomalies

Neural Computing and Applications 2024年第23期36卷 14075-14091页

作者： Astrid, Marcella Zaheer, Muhammad Zaigham Aouada, Djamila Lee, Seung-Ik Department of Artificial Intelligence University of Science and Technology Daejeon34113 Korea Republic of Field Robotics Research Section Electronics and Telecommunications Research Institute Daejeon34129 Korea Republic of Interdisciplinary Centre for Security Reliability and Trust University of Luxembourg Luxembourg1855 Luxembourg Department of Computer Vision Mohamed Bin Zayed University of Artificial Intelligence Abu Dhabi United Arab Emirates

Due to the rare occurrence of anomalous events, a typical approach to anomaly detection is to train an autoencoder (AE) with normal data only so that it learns the patterns or representations of the normal training data. At test time, the trained AE is expected to well reconstruct normal but to poorly reconstruct anomalous data. However, contrary to the expectation, anomalous data are often well reconstructed as well. In order to further separate the reconstruction quality between normal and anomalous data, we propose creating pseudo anomalies from learned adaptive noise by exploiting the aforementioned weakness of AE, i.e., reconstructing anomalies too well. The generated noise is added to the normal data to create pseudo anomalies. Extensive experiments on Ped2, Avenue, ShanghaiTech, CIFAR-10, and KDDCUP datasets demonstrate the effectiveness and generic applicability of our approach in improving the discriminative capability of AEs for anomaly detection. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Anomaly detection

来源：评论

学校读者我要写书评

暂无评论

Probabilistic Health Monitoring and Maintenance Scheduling for Nuclear Power Systems

Probabilistic Health Monitoring and Maintenance Scheduling f...

引用

AIAA Science and Technology Forum and Exposition, AIAA SciTech Forum 2025

作者： Amer, Ahmad Tsilifis, Panagiotis Saxena, Abhinav Baglietto, Emilio Khan, Genghis Probabilistic Design & Material Informatics GE Aerospace Research Niskayuna NY12309 United States Robotics AI & Computer Vision GE Vernova Advanced Research Niskayuna NY12309 United States AI & Computer Vision GE Aerospace Research San RamonCA94583 United States Department of Nuclear Science & Engineering Massachusetts Institute of Technology CambridgeMA02139 United States Sustainable Systems GE Aerospace Research Niskayuna NY12309 United States

ISBN: (数字)9781624107238

ISBN: (纸本)9781624107238

Operations and Maintenance (O&M) cost optimization in the nuclear energy industry is an imperative task for developing sustainable systems and efficient renewable technologies. We present a modular probabilistic framework for addressing maintenance scheduling challenges for small modular nuclear reactors using a Dynamic Bayesian Network (DBN) approach that monitors fatigue and performance in multiple subsystems of the reactors and provides a probabilistic assessment for potential failure events. The DBN framework fuses information from potential digital twins that assimilate real time data from several sensors with different time granularities in order to monitor and update thecurrentstate of knowledge aboutthestatus of several fatigue or performance criteria. Comprised of multi-modal digital twins (empirical, data- and physics-based twins), this DBN serves as a predictive framework forecasting the future state of specific reactor subsystems and enhancing decision-making capabilities associated with preventing system failure and minimizing the time and cost of maintenance events. The work presented in this paper is a holistic approach for monitoring and safeguarding complex, multi-component engineering systems where maintenance scheduling is critical and does not restrict itself to nuclear environments. However, our goal is to illustrate how such a framework can be adjusted to small modular water reactors. © 2025, American institute of Aeronautics and Astronautics Inc, AIAA. All rights reserved.

关键词： Nuclear energy

来源：评论

学校读者我要写书评

暂无评论

Underwater Acoustic Localization using pose-graph SLAM

Underwater Acoustic Localization using pose-graph SLAM

引用

OCEANS

作者： Marta Real Pau Vial Narcís Palomeras Marc Carreras Computer Vision and Robotics Research Institute (VICOROB) University of Girona Girona Spain

Current acoustic localization systems for Autonomous Underwater Vehicles (AUV) rely on acoustic beacons at known positions. We present a method that combines rangeonly acoustic measures and odometry data into a factor graph structure, able to compute the AUV localization while estimating the position of the acoustic beacons without having previous information about their position. We designed several missions and carried them out in a real environment to gather data to test our algorithm offline. The output of this work is a pose-graph SLAM algorithm that combines the AUV internal measurements (i.e. DVL, IMU, and depth) with acoustic ranges from static beacons to estimate both the AUV state and the beacon’s position.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Inertial navigation framework for multimodal underwater Graph SLAM

Inertial navigation framework for multimodal underwater Grap...

引用

OCEANS

作者： Pau Vial Miguel Castillón Narcís Palomeras Marc Carreras Pere Ridao Computer Vision and Robotics Research Institute (VICOROB) University of Girona Girona Spain

Robot localization is a fundamental task in achieving true autonomy for Autonomous Underwater Vehicles (AUV). If inertial measurements from an Inertial Measurement Unit (IMU) or a Doppler Velocity Log (DVL) want to be fused with some perception system, such us a multibeam sonar or several acoustic beacons; a full Simultaneous Localization And Mapping (SLAM) problem must be solved. In contrast to filters, in a full SLAM problem the whole robot trajectory is estimated and loop closure events can be detected and closed along it. Common Inertial Navigation Systems (INS), based on filters, only maintain the estimation of the current robot pose. Therefore, these systems cannot be directly used in a full SLAM problem. In this paper we present a graph solution to integrate all inertial measurements in a factor graph that can be extended to different perception modalities and it is solved by applying Smoothing and Mapping (SAM) [1]. The Preintegrated IMU factor, proposed by [2], is combined with priors for other inertial measurements that have been specially designed. This framework is tested on real data from sea experiments, showing how our proposal performance is similar to the estimation provided by high grade commercial INS products based on filters. However, our system has the advantage of allowing for fusion with exteroceptive sensors in SLAM.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Single 2D Pose with Context is Worth Hundreds for 3D Human Pose Estimation 37

A Single 2D Pose with Context is Worth Hundreds for 3D Human...

引用

37th Conference on Neural Information Processing Systems, NeurIPS 2023

作者： Zhao, Qitao Zheng, Ce Liu, Mengyuan Chen, Chen Robotics Institute Carnegie Mellon University United States Center for Research in Computer Vision University of Central Florida United States Key Laboratory of Machine Perception Peking University Shenzhen Graduate School China

ISBN: (纸本)9781713899921

The dominant paradigm in 3D human pose estimation that lifts a 2D pose sequence to 3D heavily relies on long-term temporal clues (i.e., using a daunting number of video frames) for improved accuracy, which incurs performance saturation, intractable computation and the non-causal problem. This can be attributed to their inherent inability to perceive spatial context as plain 2D joint coordinates carry no visual cues. To address this issue, we propose a straightforward yet powerful solution: leveraging the readily available intermediate visual representations produced by off-the-shelf (pre-trained) 2D pose detectors - no finetuning on the 3D task is even needed. The key observation is that, while the pose detector learns to localize 2D joints, such representations (e.g., feature maps) implicitly encode the joint-centric spatial context thanks to the regional operations in backbone networks. We design a simple baseline named Context-Aware PoseFormer to showcase its effectiveness. Without access to any temporal information, the proposed method significantly outperforms its context-agnostic counterpart, PoseFormer [74], and other state-of-the-art methods using up to hundreds of video frames regarding both speed and precision. Project page: ***/ContextAware-PoseFormer. © 2023 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

GMM Registration: a Probabilistic scan matching approach for sonar-based AUV navigation

GMM Registration: a Probabilistic scan matching approach for...

引用

IEEE International Conference on robotics and Automation (ICRA)

作者： Pau Vial Miguel Malagón Ricard Segura Narcís Palomeras Marc Carreras Computer Vision and Robotics Research Institute (VICOROB) Universitat de Girona Girona Catalonia

Acoustic perception in underwater environments is challenging due to the low frequency of the acquisition system and multiple and huge sources of noise. Therefore, point clouds built by profiling sonars mounted on Autonomous Underwater Vehicles (AUV) are sparse and noisy. To solve the mapping task, AUVs need a registration algorithm to prevent maps from inconsistencies. Many scan matching algorithms are available, however, a few of them are specialized in acoustic data. In this paper, a probabilistic scan matching methodology based on Gaussian Mixtures Models (GMM) is presented and, for the first time, the Bayesian-GMM algorithm is applied in this context to model acoustic data. The scan matching problem is properly formulated using Lie groups to define pose. In addition, this methodology can return an uncertainty measure for the matching result, which is fundamental in Pose SLAM applications. This tool is implemented in a public C++library 1 1 The library repository can be found in https://***/gmmregistration/gmm_registration. that can process in real-time 2D and 3D scans acquired by a profiling sonar. Theoretical justification and results with real data are provided to benchmark our method against the state-of-the-art Normal Distributions Transforms (NDT) technique.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Frequency Guidance Matters: Skeletal Action Recognition by Frequency-Aware Mixed Transformer 24

Frequency Guidance Matters: Skeletal Action Recognition by F...

引用

32nd ACM International Conference on Multimedia, MM 2024

作者： Wu, Wenhan Zheng, Ce Yang, Zihao Chen, Chen Das, Srijan Lu, Aidong Department of Computer Science University of North Carolina at Charlotte CharlotteNC United States Robotics Institute Carnegie Mellon University PittsburghPA United States One Inventory Organization Microsoft Corporation RedmondWA United States Center for Research in Computer Vision University of Central Florida OrlandoFL United States

ISBN: (纸本)9798400706868

Recently, transformers have demonstrated great potential for modeling long-term dependencies from skeleton sequences and thereby gained ever-increasing attention in skeleton action recognition. However, the existing transformer-based approaches heavily rely on the naive attention mechanism for capturing the spatiotemporal features, which falls short in learning discriminative representations that exhibit similar motion patterns. To address this challenge, we introduce the Frequency-aware Mixed Transformer (FreqMixFormer), specifically designed for recognizing similar skeletal actions with subtle discriminative motions. First, we introduce a frequency-aware attention module to unweave skeleton frequency representations by embedding joint features into frequency attention maps, aiming to distinguish the discriminative movements based on their frequency coefficients. Subsequently, we develop a mixed transformer architecture to incorporate spatial features with frequency features to model the comprehensive frequency-spatial patterns. Additionally, a temporal transformer is proposed to extract the global correlations across frames. Extensive experiments show that FreqMiXFormer outperforms SOTA on 3 popular skeleton action recognition datasets, including NTU RGB+D, NTU RGB+D 120, and NW-UCLA datasets. Our project is publicly available at: https://***/wenhanwu95/FreqMixFormer. © 2024 ACM.

关键词： Musculoskeletal system

来源：评论

学校读者我要写书评

暂无评论

Complexity based investigation in collaborative assembly scenarios via non intrusive techniques 4

Complexity based investigation in collaborative assembly sce...

引用

4th International Conference on Industry 4.0 and Smart Manufacturing, ISM 2022

作者： Panagou, Sotirios Sileo, Monica Papoutsakis, Konstantinos Fruggiero, Fabio Qammaz, Ammar Argyros, Antonis School of Engineering University of Basilicata Potenza85100 Italy Computational Vision and Robotics Laboratory Institute of Computer Science Foundation for Research and Technology - Hellas Heraklion70013 Greece

Human and robot collaboration in assembly tasks is an integral part in modern manufactories. Robots provide advantages in both process and productivity with their repeatability and usability in different tasks, while human operators provide flexibility and can act as safeguards. However, process complexity increases which can lower the overall quality. Increased complexity can negatively influence decision making due to cognitive load on human operators, which can lead to lower quality, be it product, process or human work. Moreover, it can lead to safety risks, human-system error and accidents. In this work, we present the preliminary results on an experiment performed with student-participants, based on an assembly task. The experiment was set up to emulate an industrial assembly, and data collection was performed through qualitative and non-intrusive quantitative methods. Questionnaires were used to assess perceptual task complexity and cognitive load, while a stereo camera provided recordings for after-task analysis on process errors and human work quality based on a 3D skeleton-based human pose estimation and tracking method. The aim of the study is to investigate causes of errors and implications on quality. Future direction of the work is discussed. © 2022 The Authors. Published by ELSEVIER B.V.

关键词： Decision making

来源：评论

学校读者我要写书评

暂无评论

Mitigating Overdiagnosis Bias in CNN-Based Alzheimer’s Disease Diagnosis for the Elderly 2nd

Mitigating Overdiagnosis Bias in CNN-Based Alzheimer’s Dis...

引用

2nd International Workshop on Fairness of AI in Medical Imaging, FAIMI 2024, and 3rd International Workshop on Ethical and Philosophical Issues in Medical Imaging, EPIMI 2024, Held in Conjunction with the International Conference on Medical Image Computing and computer Assisted Interventions, MICCAI 2024

作者： Dang, Vien Ngoc Casamitjana, Adrià Hernández-González, Jerónimo Lekadir, Karim Departament de Matemàtiques i Informàtica Universitat de Barcelona Barcelona Spain Research Institute of Computer Vision and Robotics University of Girona Girona Spain Departament d’Informàtica Matemàtica Aplicada i Estadística Universitat de Girona Girona Spain Barcelona Spain

ISBN: (纸本)9783031727863

Diagnosing Alzheimer’s disease (AD) presents significant challenges in the oldest populations due to overlapping symptoms of normal cognitive aging and early-stage dementia. While AI algorithms have matched specialist performance in diagnosing AD, they tend to produce unreliable results for the oldest populations, generating false positives that increase radiologist workloads and healthcare costs. In this study, we focus on mitigating overdiagnosis bias in CNN-based AD diagnosis for these groups. We present a post-hoc bias mitigation technique that significantly improves fairness by reducing overdiagnosis and enhances reliability by improving calibration without compromising overall model accuracy. Code is available at: https://***/ngoc-vien-dang/C-GTOP. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Robotic Approach Trajectory Using Reinforcement Learning with Dual Quaternions

Robotic Approach Trajectory Using Reinforcement Learning wit...

引用

Iberian robotics Conference (ROBOT)

作者： Daniel Frau-Alfaro Santiago T. Puente Ignacio De Loyola Páez-Ubieta Edison Velasco-Sánchez AUtomatics RObotics and Artificial Vision Lab University Institute for Computer Research San Vicente Spain University Institute for Engineering Research University Miguel Hernández Elche Spain

ISBN: (数字)9798350376364

ISBN: (纸本)9798350376371

Manipulation tasks in robotics usually involve two phases: an approach to the object and the grasp itself. The first action allows the robot to reach a certain pose in space that is likely to allow the object to be manipulated. Reinforcement Learning (RL) techniques allow a policy to be learned through experience given a set of states and actions, so this is a powerful tool for developing controllers for specific tasks such as positioning the robot in a particular point in space. However, when manipulating an object, orientation is as relevant as position. For this reason, a method of RL for positioning the robot's end effector in a suitable position and orientation for manipulation in simulation is presented. This approach models the problem of computing the distance for the reward function using dual quaternions parameterisation, an element that can represent the pose and attitude of a rigid body in Euclidean space in a compact way without having to apply any constraints.

关键词： Training Quaternions Computational modeling Process control Reinforcement learning Aerospace electronics End effectors Trajectory Robots

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：