检索结果-内蒙古大学图书馆

Black-box Explainability-Guided Adversarial Attack For 3D Object Tracking

IEEE Transactions on Circuits and Systems for Video technology 2025年

作者： Cheng, Riran Wang, Xupeng Sohel, Ferdous Lei, Hang University of Electronic Science and Technology of China School of Information and Software Engineering Sichuan Chengdu China Murdoch University School of Information Technology Perth Australia

With the development of deep 3D tracking models and their broad prospects for safety-critical applications, adversarial robustness, i.e., the ability of deep models to resist malicious adversarial attacks, has become an important research topic. Previous works generate adversarial examples by tampering with points of the input point cloud indiscriminately. Consequently, they suffer from high computing costs and limited attack performance caused by the trade-off between imperceptibility and adversarial strength. In this paper, we propose a novel adversarial attack against 3D object tracking, which is guided by an occlusion-based explainability method to target points crucial for the predictions in the search area and results in a significant deviation between the predictions and the ground truth. Specifically, an attribution map is generated to reveal the importance of points to the model decision, which is achieved by measuring the variations of tracking performance under subsets generated by the downsampling strategy. To facilitate the generation of attribution maps, the downsampling strategy considers prior knowledge of 3D trackers, which assigns higher sampling probabilities to points with potentially higher contributions enclosed by bounding boxes. Multi-scale fusion is also leveraged to integrate the sensitivity of the model to local regions of varying sizes. Considering the requirement of imperceptibility on adversarial attacks, a hard geometric constraint is imposed on the targeted critical points, which produces perturbations with the property of surface invariance. Furthermore, in contrast to existing works devoted to spatial information manipulation only, multiple loss functions are developed to guide the perturbation generation, where the predicted motions of the tracking target representing the spatial-temporal information unique to the tracking task are distorted to deceive 3D trackers. Extensive experiments conducted on public benchmarks and 3D tracker

关键词： 3D object tracking adversarial attack attribution map explainability imperceptibility

来源：评论

学校读者我要写书评

暂无评论

Heterogeneous signcryption scheme from CLC to IBC for IIoT

引用

Peer-to-Peer Networking and Applications 2025年第4期18卷 1-14页

作者： Jin, Chunhua Qin, Wenyu Chen, Zhiwei Li, Chenhao Chen, Xiaobing Chen, Guanhua Zhang, Hao Weng, Jian Faculty of Computer and Software Engineering Huaiyin Institute of Technology Huai’an China The College of Information Science and Technology Jinan University Zhuhuai China

The Industrial Internet of Things (IIoT) involves the real-time gathering of information from physical devices and technologies for improved perception and management. To ensure information transmission security, an efficient and secure cryptography scheme is needed, especially for devices under resource constraints. In this paper, a heterogeneous signcryption (HSC) scheme is designed to meet security requirements in the IIoT environment. This construction employs a heterogeneous system, with the sender using certificateless cryptography (CLC) to effectively solve the key escrow problem in identity-based cryptography (IBC) and the certificate management problem in public key infrastructure (PKI). The receiver, on the other hand, uses IBC to settle the certificate management issue in PKI. The scheme supports formal security proof in the random oracle model (ROM). Furthermore, the detailed performance analysis demonstrates that the scheme has advantages in terms of energy consumption and computation overhead, making it suitable for the IIoT environment.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Comprehensive Review on Deep Learning System Testing 24th

A Comprehensive Review on Deep Learning System Testing

引用

24th International Conference on Algorithms and Architectures for Parallel Processing, ICA3PP 2024

作者： Li, Ying Shan, Chun Liu, Zhen Liao, Shuyan Beijing Key Laboratory of Software Security Engineering Technology Beijing Institute of Technology Beijing China School of Computer Science and Technology Beijing Institute of Technology Beijing China Computer School Beijing Information Science and Technology University Beijing China School of Cyberspace Science and Technology Beijing Institute of Technology Beijing China

ISBN: (纸本)9789819615476

Deep learning(DL) systems exhibit multiple behavioral characteristics such as correctness, robustness, and fairness. Ensuring that these behavioral characteristics function properly is crucial for maintaining the accuracy of DL systems’ outputs. As a specialized form of software, DL systems’ security testing techniques have increasingly become a focus of research in quality assurance. We analyze and organize the testing techniques for DL systems based on an investigation of the current state of the art both domestically and internationally. This paper categorizes existing approaches as component-oriented and attribute-oriented methods, providing a detailed review based on this classification. Additionally, we forecast the future development of testing techniques for DL systems. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Hierarchical spatial-spectral enhancement network for hyperspectral image and light detection and ranging data classification

引用

Journal of Applied Remote Sensing 2025年第1期19卷

作者： Yang, Zhichao Hao, Siyuan Li, Enyu Zhao, Kun Qingdao University of Technology College of Information and Control Engineering Qingdao China Beijing Jiaotong University College of Software Engineering Beijing China

The effective utilization of multimodal data is currently a prominent research focus in remote sensing. Hyperspectral image (HSI) and light detection and ranging (LiDAR) image can provide unique and complementary information. HSI can capture the spectral reflection characteristics of ground objects across hundreds of contiguous bands, whereas LiDAR can accurately measure the elevation of ground surfaces, providing precise information on topography and landforms. Combining them can yield more comprehensive and accurate information about ground objects. However, due to the underutilization of complementary information from multisource datasets, the extraction and integration of spatial and spectral information remain insufficient after fusion. We developed a hierarchical spatial-spectral enhancement network (HSSE-Net) for the classification of HSI and LiDAR data. Specifically, by employing convolutional operations to capture the spatial features of both HSI and LiDAR, we enhance the spatial feature expression capability of each modality. Next, we use a dual-branch network integrated with a transformer self-attention mechanism to process the fused data, which helps capture long-range dependencies and global information in the spectral data. This hierarchical enhancement of spatial and spectral information can further improve the cooperative classification performance. Extensive experiments conducted on two popular HSI and LiDAR datasets show that the HSSE-Net achieves average overall classification accuracies of 99.26% and 94.96% on the two datasets, respectively, fully demonstrating the effectiveness and superior performance of the proposed method. © 2025 Society of Photo-Optical Instrumentation Engineers (SPIE).

关键词： Photointerpretation

来源：评论

学校读者我要写书评

暂无评论

RETHINKING SPIKING NEURAL NETWORKS FROM AN ENSEMBLE LEARNING PERSPECTIVE

arXiv

引用

arXiv 2025年

作者： Ding, Yongqi Zuo, Lin Jing, Mengmeng He, Pei Deng, Hanpu School of Information and Software Engineering University of Electronic Science and Technology of China China

Spiking neural networks (SNNs) exhibit superior energy efficiency but suffer from limited performance. In this paper, we consider SNNs as ensembles of temporal subnetworks that share architectures and weights, and highlight a crucial issue that affects their performance: excessive differences in initial states (neuronal membrane potentials) across timesteps lead to unstable subnetwork outputs, resulting in degraded performance. To mitigate this, we promote the consistency of the initial membrane potential distribution and output through membrane potential smoothing and temporally adjacent subnetwork guidance, respectively, to improve overall stability and performance. Moreover, membrane potential smoothing facilitates forward propagation of information and backward propagation of gradients, mitigating the notorious temporal gradient vanishing problem. Our method requires only minimal modification of the spiking neurons without adapting the network structure, making our method generalizable and showing consistent performance gains in 1D speech, 2D object, and 3D point cloud recognition tasks. In particular, on the challenging CIFAR10-DVS dataset, we achieved 83.20% accuracy with only four timesteps. This provides valuable insights into unleashing the potential of SNNs. © 2025, CC BY-NC-ND.

关键词： Spatio-temporal data

来源：评论

学校读者我要写书评

暂无评论

An Experimental Study on Dynamic Lifelong Learning with GPT for Mitigating Catastrophic Forgetting in Aspect-Based Sentiment Analysis

引用

IEEE Access 2025年

作者： Huang, Huang Asemi, Adeleh Universiti Malaya Faculty of Computer Science Information Technology Department of Software Engineering Kuala Lumpur Malaysia

GPT is widely recognized as one of the most versatile and powerful large language models, excelling across diverse domains. However, its significant computational demands often render it economically unfeasible for individuals and small businesses, underscoring the need for efficient, domainspecific alternatives. In Aspect-Based Sentiment Analysis (ABSA), existing models are typically optimized for single domains, facing challenges in performing effectively across multiple domains. A key issue, known as "catastrophic forgetting," arises when models trained on one domain lose previously learned knowledge upon exposure to new domain data. This leads to two significant problems: limited cross-domain generalization and difficulty in retaining prior knowledge while learning domain-specific information. To address these challenges, we introduce the Dynamic Lifelong Learning Aspect-Based Sentiment Analysis GPT (DllaGPT), a model designed to handle multiple domains while mitigating catastrophic forgetting. Leveraging datasets from four ABSA domains—Laptops, Restaurants, Tweets, and Finance—this study fine-tunes a pretrained GPT model from HuggingFace sequentially across domains. DllaGPT employs a mechanism to retain real data from earlier domains during new domain training, effectively preserving prior knowledge. Experimental results highlight that DllaGPT achieves an average accuracy of 0.85 and a Backward Transfer (BWT) score of -0.09 across the four domains, showcasing its high accuracy and robust lifelong learning capabilities. © 2025 Institute of Electrical and Electronics Engineers Inc.. All rights reserved.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

A Multi-feature Co-distillation Algorithm Based on Human Skeletal Point Behavior Recognition 24

A Multi-feature Co-distillation Algorithm Based on Human Ske...

引用

2nd International Conference on Advances in Artificial Intelligence and Applications, AAIA 2024

作者： Pan, Xuefeng Tian, Zhiqiang School of Computer Wuhan Qingchuan University Wuhan China School of Information Wuhan Software Engineering Vocational College Wuhan Open University Wuhan China School of Computer Science and Technology Wuhan University of Science and Technology Wuhan China

ISBN: (纸本)9798400712883

Depth model-based behavior estimation of human skeletal points is widely used in the field of behavior recognition. In order to improve the accuracy of behavior recognition, the complexity and computation of the designed network model are increasing, while the mobile platforms have limited arithmetic power and memory, and thus cannot be deployed. To solve the above problems, the AGCN network, a skeletal point-based behavior recognition model, is clipped to compress the network size so that it can support common mobile platforms;and a multi-feature co-distillation method is proposed to efficiently migrate the performance of the original network to the target network, so that the performance of the clipped lightweight network tends to be close to that of the original network. The model training, testing, visualization, and ablation experiments are completed on the NTU-RGB+D60 data set. The experimental results show that the algorithm after lightweighting (clipping network and multi-feature co-distillation), can significantly reduce the computational amount of behavior recognition. © 2024 Copyright held by the owner/author(s).

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Virtual Reality Human-Computer Interaction System Based on Artificial Intelligence technology 14th

Virtual Reality Human-Computer Interaction System Based on A...

引用

14th International Conference on Frontier Computing, FC 2024

作者： Li, Jiaqi Xiong, Gaojie Wu, Xing Guangdong Shenzhen518000 China School of Information and Software Engineering University of Electronic Science and Technology of China Sichuan Chengdu610000 China Zhejiang Vocational Academy of Art Zhejiang Hangzhou310053 China

ISBN: (纸本)9789819623907

Although artificial intelligence and virtual reality technology have been rapidly popularized, current human-computer interaction systems still have challenges such as network latency, data synchronization, and natural communication limitations. This article aims to explore ways to use artificial intelligence technology to improve human-computer interaction in virtual reality, evaluate its effects, and introduce the development status of artificial intelligence technology and virtual reality technology and their applications in the field of human-computer interaction. This article first analyzes the needs and challenges of virtual reality human-computer interaction systems, and then proposes an improvement method based on artificial intelligence technology, including using machine learning algorithms to analyze and predict user behavior, and using natural language processing technology to achieve natural language interaction. The research results show that the reliability of this system is up to 96%. Virtual reality human-computer interaction systems based on artificial intelligence technology have made significant improvements in improving user experience and interaction effects. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Virtual environments

来源：评论

学校读者我要写书评

暂无评论

Point2Quad: Generating Quad Meshes from Point Clouds via Face Prediction

引用

IEEE Transactions on Circuits and Systems for Video technology 2025年

作者： Li, Zezeng Qi, Zhihui Wang, Weimin Wang, Ziliang Duan, Junyi Lei, Na Dalian University of Technology School of Software Technology DUT-RU International School of Information Science and Engineering Dalian116024 China

Quad meshes are essential in geometric modeling and computational mechanics. Although learning-based methods for triangle mesh demonstrate considerable advancements, quad mesh generation remains less explored due to the challenge of ensuring coplanarity, convexity, and quad-only meshes. In this paper, we present Point2Quad, the first learning-based method for quad-only mesh generation from point clouds. The key idea is learning to identify quad mesh with fused pointwise and facewise features. Specifically, Point2Quad begins with a k-NN-based candidate generation considering the coplanarity and squareness. Then, two encoders are followed to extract geometric and topological features that address the challenge of quad-related constraints, especially by combining in-depth quadrilaterals-specific characteristics. Subsequently, the extracted features are fused to train the classifier with a designed compound loss. The final results are derived after the refinement by a quad-specific post-processing. Extensive experiments on both clear and noise data demonstrate the effectiveness and superiority of Point2Quad, compared to baseline methods under comprehensive metrics. The code and dataset will be publicly available. © 1991-2012 IEEE.

关键词： Mesh generation

来源：评论

学校读者我要写书评

暂无评论

Design and Modeling of a Photonic Crystal Multiplexer Using Artificial Intelligence

Advanced Electromagnetics

引用

Advanced Electromagnetics 2025年第1期14卷 59-64页

作者： Karami, Pouya Yahya, Salah I. Palash, Behnam Chaudhary, Muhammad Akmal Assaad, Maher Parandin, Fariborz Roshani, Saeed Hazzazi, Fawwaz Roshani, Sobhan Department of Electrical Engineering Kermanshah Branch Islamic Azad University Kermanshah Iran Department of Communication and Computer Engineering Cihan University-Erbil Erbil44001 Iraq Department of Software Engineering Faculty of Engineering Koya University KOY45 Koya Iraq Department of Electrical and Computer Engineering College of Engineering and Information Technology Ajman University 346 Ajman United Arab Emirates Department of Electrical Engineering College of Engineering Prince Sattam bin Abdulaziz University Al-Kharj11492 Saudi Arabia

In this paper, design and modeling of an all-optical 2×1 multiplexer based on 2D photonic crystals and artificial neural networks (ANNs) are presented. The proposed structure aims to maximize the difference between the output powers in logical states, which is critical for enhancing the system ability to distinguish between these states. In this study, an ANN model is employed to accurately predict the normalized output power of the designed photonic crystal multiplexer, providing a time-efficient alternative to conventional simulation methods for analyzing multiplexer behavior across various logical states. The results demonstrate significant improvements in signal separation and overall performance compared to previous works. Additionally, a detailed comparison of the normalized output power for different logic states is provided, highlighting the advantages of the proposed design. © 2025, Advanced Electromagnetics. All rights reserved.

关键词： Silicon photonics

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：