检索结果-内蒙古大学图书馆

RGB-Based Category-Level Object Pose Estimation via Depth Recovery and Adaptive Refinement

IEEE Robotics and Automation Letters 2025年第6期10卷 5377-5384页

作者： Yang, Hui Sun, Wei Liu, Jian Zheng, Jin Zeng, Zhiwen Mian, Ajmal Hunan University National Engineering Research Center for Robot Visual Perception and Control Technology College of Electrical and Information Engineering Changsha410082 China Hunan University State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body Changsha410082 China Central South University School of Architecture and Art Changsha410082 China University of Western Australia Department of Computer Science and Software Engineering CrawleyWA6009 Australia

Category-level pose estimation methods have received widespread attention as they can be generalized to intra-class unseen objects. Although RGB-D-based category-level methods have made significant progress, reliance on depth image limits practical application. RGB-based methods offer a more practical and cost-effective solution. However, current RGB-based methods struggle with object geometry perception, leading to inaccurate pose estimation. We propose depth recovery and adaptive refinement for category-level object pose estimation from a single RGB image. We leverage DINOv2 to reconstruct the coarse scene-level depth from the input RGB image and propose an adaptive refinement network based on an encoder-decoder architecture to dynamically improve the predicted coarse depth and reduce its gap from the ground truth. We introduce a 2D–3D consistency loss to ensure correspondence between the point cloud obtained from depth projection and the objects in the 2D image. This consistency supervision enables the model to maintain alignment between the depth image and the point cloud. Finally, we extract features from the refined point cloud and feed them into two confidence-aware rotation regression branches and a translation and size prediction residual branch for end-to-end training. Decoupling the rotation matrix provides a more direct representation, which facilitates parameter optimization and gradient propagation. Extensive experiments on the REAL275 and CAMERA25 datasets demonstrate the superior performance of our method. Real-world estimation and robotic grasping experiments demonstrate our model robustness to occlusion, clutter environments, and low-textured objects. Our code and robotic grasping video are available at DA-Pose. © 2016 IEEE.

关键词： Depth perception

来源：评论

学校读者我要写书评

暂无评论

Unsupervised Distributed Anomaly Detection Framework for IoT in Edge AI Network

引用

IEEE Internet of Things Journal 2025年

作者： Jiang, Xingguo Lu, Chang Luo, Hong Sun, Yan Beijing University of Posts and Telecommunications School of Computer Science National Pilot Software Engineering School Beijing100876 China

Anomaly detection of sensor data is crucial to ensure the stability and effectiveness of IoT system. The task requires high accuracy and low latency, which makes distributed anomaly detection gradually become a research hotspot. Edge AI networks further enhance the computing power of edge servers, allowing distributed anomaly detection to be gradually applied to complex scenarios under the Industrial Internet of Things (IIoT), such as smart factories. However, in such cases, the data reported by different sensors in different fields are not independent and identically distributed (non-IID). Simultaneously, during unsupervised training, anomaly data mixed into the training data will reduce the recognition ability of the anomaly detector. To address these challenges, we propose an unsupervised distributed anomaly detection framework. On the cloud, we train an unsupervised anomaly detection model with global factors, using global and local factors to learn the distribution patterns of different fields. At the edge, a multi-dimensional threshold and its automatic selection algorithm are proposed to overcome the problem of decreased anomaly recognition ability introduced by anomaly training data. Extensive experiments on six datasets show that our approach outperforms SOTA methods in F1-score, can detect anomalies with high accuracy and efficiency in distributed IoT scenarios. © 2014 IEEE.

关键词： Industrial internet of things (IIoT)

来源：评论

学校读者我要写书评

暂无评论

HTM-CDFK: An Online Industrial Control Anomaly Detection Algorithm based on Hierarchical Time Memory

引用

IEEE Transactions on Dependable and Secure Computing 2025年

作者： Liu, Jingwen Zhang, Yuanyuan Wang, Zilin Ren, Yi Chen, Liangyin Chen, Yanru Sichuan University College of Computer Science College of Software Engineering Institute for Industrial Internet Research Chengdu610065 China

With the rapid advancement of industrial structures, many factories are deploying anomaly detection systems. However, most existing anomaly detection algorithms are unsuitable for high-noise industrial control system (ICS) environments and constantly changing sensor patterns, resulting in low detection accuracy. To address these challenges, we propose a hierarchical temporal memory-based online anomaly detection algorithm for ICS, incorporating cumulative distribution functions and Gaussian convolution kernels. We enhanced the encoding process of hierarchical temporal memory, enabling it to quickly fit the physical processes of ICS and remember the system's previous operating states for anomaly recognition. Additionally, our anomaly calculation algorithm, based on cumulative distribution functions and Gaussian kernel convolution, effectively addresses the adaptability and accuracy challenges posed by high-noise ICS environments. Experimental results demonstrate that our method outperforms the best baseline algorithms. While maintaining good time stability, our method achieves a 9% improvement in detection accuracy, highlighting its significant advantage in balancing detection precision and time performance. © 2004-2012 IEEE.

关键词： Distribution functions

来源：评论

学校读者我要写书评

暂无评论

Flexible ViG: Learning the Self-Saliency for Flexible Object Recognition

引用

IEEE Transactions on Circuits and Systems for Video Technology 2025年

作者： Yang, Kunshan Zuo, Lin Jing, Mengmeng Tian, Xianlong He, Kunbin Ding, Yongqi University of Electronic Science and Technology of China School of Information and Software Engineering ChengDu611731 China University of Electronic Science and Technology of China School of Computer Science and Engineering ChengDu611731 China

Existing computer vision methods mainly focus on the recognition of rigid objects, whereas the recognition of flexible objects remains unexplored. Recognizing flexible objects poses significant challenges due to their inherently diverse shapes and sizes, translucent attributes, ambiguous boundaries, and subtle inter-class differences. In this paper, we claim that these problems primarily arise from the lack of object saliency. To this end, we propose the Flexible Vision Graph Neural Network (FViG) to optimize the self-saliency and thereby improve the discrimination of the representations for flexible objects. Specifically, on one hand, we propose to maximize the channel-aware saliency by extracting the weight of neighboring graph nodes, which is employed to identify flexible objects with minimal inter-class differences. On the other hand, we maximize the spatial-aware saliency based on clustering to aggregate neighborhood information for the centroid graph nodes. This introduces local context information and enables extracting of consistent representation, effectively adapting to the shape and size variations in flexible objects. To verify the performance of flexible objects recognition thoroughly, for the first time we propose the Flexible Dataset (FDA), which consists of various images of flexible objects collected from real-world scenarios or online. Extensive experiments evaluated on our FDA, FireNet, CIFAR-100 and ImageNet-Hard datasets demonstrate the effectiveness of our method on enhancing the discrimination of flexible objects. © 1991-2012 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

A Meta-Model to Support Compatibility Testing of Cross-Browser Web Application

A Meta-Model to Support Compatibility Testing of Cross-Brows...

引用

International Conference on Advancements in Computational sciences (ICACS)

作者： Mishal Muneer Uzair Rasheed Muhammad Majid Hussain Department of Computer Science & IT The University of Lahore Lahore Pakistan Department of Computer and Software Engineering College of Electrical and Mechanical Engineering NUST Islamabad Pakistan

ISBN: (数字)9798331533038

ISBN: (纸本)9798331533045

Ensuring the user interface (UI) compatibility of web applications across diverse client-side configurations, including various operating systems and browsers, is a significant challenge due to the extensive range of possible combinations. Traditional tools often struggle to address this complexity, leading to visual inconsistencies and malfunctions. To improve compatibility testing, we propose a meta-model that initiates the process with a checklist covering critical configurations across different browsers. This checklist is then translated into Interaction Flow Modeling Language (IFML) constructs, enabling a model-driven approach to compatibility assessment. By leveraging this checklist-IFML integration, we systematically generate test cases that target compatibility issues more effectively. Our approach is validated through the Laptop-Web Case Study, where we assess compatibility by comparing checklist items with IFML constructs within the case study’s domain model. The results show that our approach enhances the accuracy and efficiency of compatibility testing, addressing a core gap in current methods. This research introduces a structured, model-driven method for compatibility testing, providing a more reliable framework for identifying and mitigating UI compatibility issues in web applications across varying client-side configurations.

关键词： Visualization Translation Flow production systems Computational modeling Operating systems User interfaces Reliability engineering Browsers Complexity theory Testing

来源：评论

学校读者我要写书评

暂无评论

Phoneme Classification in Imagined Speech Using Explainable Machine Learning 13

Phoneme Classification in Imagined Speech Using Explainable ...

引用

13th International Winter Conference on Brain-computer Interface, BCI 2025

作者： Kim, Sejin Kang, Hyunwook Jeong, Ji-Hoon Lee, Minji Dept. Artificial Intelligence The Catholic University of Korea Bucheon Korea Republic of Dept. Electrical and Computer Engineering Inha University Incheon Korea Republic of Dept. Computer Science Chungbuk National University Cheongju Korea Republic of Dept. Biomedical Software Engineering The Catholic University of Korea Bucheon Korea Republic of

ISBN: (纸本)9798331521929

Imagined speech is gaining attention as a next-generation paradigm for brain-computer interfaces in terms of its intuitiveness in communication. Many studies have focused on classifying imagined words as the basis of communication, but some studies on phoneme classification based on imagined speech have begun to be conducted in terms of the fact that combinations of phonemes can further expand sound images. In addition, explainability is a major issue in classification results in machine learning, but there have been few studies that analyzed the explainability of imagined speech classification. In this study, we examined explainability using SHAP for phoneme classification based on imagined speech. FEIS dataset was used, and 16 phonemes were classified using the XGBoost classifier. In 21 subjects, the average accuracy was 40.36%, the average precision was 41.7%, the average recall was 40.34%, and the average F1-score was 40.7%. In addition, when looking at Sub 04, which had the highest performance, the SHAP values were high in AF3 and F7 channels, which correspond to Wernicke’s area related to speech. In this regard, we could observe the feature importance of this area. These results prove that Wernicke’s area actually plays an important role in classifying phoneme units in imagined speech, and can provide important insights to improve the performance of imagined speech for intuitive communication based on brain-computer interface in the future. © 2025 IEEE.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Human–object interaction detector with unsupervised domain adaptation

引用

Knowledge-Based Systems 2025年 321卷

作者： Cheng, Yamin Duan, Zhekai Huang, Hualong Wang, Zhi School of Computer and Software Engineering Xihua University Chengdu China Department of Computer Science University College London United Kingdom School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China

Human–object interaction (HOI) detection constitutes a foundational challenge in fine-grained visual scene understanding. While mainstream HOI detection frameworks rely on supervised training of deep neural networks with extensive labeled datasets, such data-intensive paradigms face practical limitations due to the high annotation and costs in real-world applications. To tackle this problem, it is a natural idea to leverage data from a well-labeled domain (source domain) to improve the performance for the domain of interest (target domain). However, the domain shift problem renders it non-trivial to transfer knowledge between different domains. In this paper, we propose a novel Domain-Adaptive HOI Detection Framework that effectively bridges domain gaps and achieves robust performance on unlabeled target domains by leveraging knowledge from labeled source domains. Specifically, our method aims to align features on both the map-level and sequence-level to learn domain-invariant features, and minimize the interaction-aware feature discrepancy by learning prototypical patterns of action relationship. To this end, we introduce a new adversarial training schema that uses a CutMix strategy to minimize the domain shift of map-level feature representations from the feature backbone. We then delve into feature distribution alignment and pursue sequence-level domain-invariant features from the transformer encoder by designing a token-based feature alignment. In addition, we design a memory bank to minimizes the distance between each inter-class interaction feature with homogeneous relationship, and maximizes the discrepancy between inter-class interaction features with heterogenous relationship. Extensive experiments on multiple benchmarks demonstrate that our proposed framework significantly outperforms prior supervised learning methods and achieves the superior performance. © 2025 Elsevier B.V.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

The 23andMe Data Breach: Analyzing Credential Stuffing Attacks, Security Vulnerabilities, and Mitigation Strategies

arXiv

引用

arXiv 2025年

作者： Holthouse, Ryan Owens, Serena Bhunia, Suman Department of Computer Science and Software Engineering Miami University OxfordOH45056 United States

In October 2023, 23andMe, a prominent provider of personal genetic testing, ancestry, and health information services, suffered a significant data breach orchestrated by a cybercriminal known as "Golem." Initially, approximately 14,000 user accounts were compromised by a credential smear attack, exploiting reused usernames and passwords from previous data leaks. However, due to the interconnected nature of 23andMe’s DNA Relatives and Family Tree features, the breach expanded exponentially, exposing sensitive personal and genetic data of approximately 5.5 million users and 1.4 million additional profiles. The attack highlights the increasing threat of credential stuffing, exacerbated by poor password hygiene and the absence of robust security measures such as multi-factor authentication (MFA) and rate limiting. In response, 23andMe mandated password resets, implemented email-based two-step verification, and advised users to update passwords across other services. This paper critically analyzes the attack methodology, its impact on users and the company, and explores potential mitigation strategies, including enhanced authentication protocols, proactive breach detection, and improved cybersecurity practices. The findings underscore the necessity of stronger user authentication measures and corporate responsibility in safeguarding sensitive genetic and personal data. © 2025, CC BY.

关键词： Cybersecurity

来源：评论

学校读者我要写书评

暂无评论

Breaking the Vault: A Case Study of the 2022 LastPass Data Breach

arXiv

引用

arXiv 2025年

作者： Gentles, Jessica Fields, Mason Goodman, Garrett Bhunia, Suman Department of Computer Science and Software Engineering Miami University OxfordOH45056 United States

Managing the security of employee work computers has become increasingly important as today’s work model shifts to remote and hybrid work plans. In this paper, we explore the recent 2022 LastPass data breach, in which the attacker obtained sensitive customer data by exploiting a software vulnerability on a DevSecOps engineer’s computer. We discuss the methodology of the attacker as well as the impact this incident had on LastPass and its customers. Next, we expand upon the impact the breach had on LastPass as well as its customers. From this, we propose solutions for preparing for and mitigating similar attacks in the future. The aim of this paper is to shed light on the LastPass incident and provide methods for companies to secure their employee base, both nationally and internationally. With a strong security structure, companies can vastly reduce the chances of falling victim to a similar attack. © 2025, CC BY.

关键词： Cloud security

来源：评论

学校读者我要写书评

暂无评论

Efficient and Effective Prompt Tuning via Prompt Decomposition and Compressed Outer Product

arXiv

引用

arXiv 2025年

作者： Lan, Pengxiang Xu, Haoyu Yang, Enneng Liang, Yuliang Guo, Guibing Zhao, Jianzhe Wang, Xingwei Software College Northeastern University China School of Computer Science and Engineering Northeastern University China

Prompt tuning (PT) offers a cost-effective alternative to fine-tuning large-scale pre-trained language models (PLMs), requiring only a few parameters in soft prompt tokens added before the input text. However, existing PT approaches face two significant issues: (i) They overlook intrinsic semantic associations between soft prompt tokens, leading to high discreteness and limited interactions, thus reducing the model’s comprehension and effectiveness in complex tasks. (ii) Due to the complexity of downstream tasks, long soft prompt is necessitated to improve performance, but prompt length correlates positively with memory usage and computational costs. Achieving high efficiency and performance remains an ongoing challenge. To address these issues, we propose a novel Low-pAraMeters Prompt Tuning (LAMP) method, which leverages prompt decomposition and compressed outer product. Specifically, the prompt decomposition module employs Truncated SVD to reduce training parameters and significantly lower the dimensionality of the soft prompt parameter space. It then utilizes a compressed outer product module to facilitate multiple interactions among prompt tokens, exploring their intrinsic associations to enhance knowledge representation. Finally, LAMP uses average pooling to reduce memory usage and training/inference time. Extensive experiments across six architectures and eight datasets demonstrate that LAMP outperforms state-of-the-art PT-based and LoRA-based methods in performance and efficiency. © 2025, CC BY.

关键词： Semantics

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：