检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Li, Depeng Zeng, Zhigang The School of Artificial Intelligence and Automation The Institute of Artificial Intelligence Huazhong University of Science and Technology China The Key Laboratory of Image Processing and Intelligent Control of Education Ministry of China Wuhan430074 China

In the scenario of class-incremental learning (CIL), deep neural networks have to adapt their model parameters to non-stationary data distributions, e.g., the emergence of new classes over time. However, CIL models are challenged by the well-known catastrophic forgetting phenomenon. Typical methods such as rehearsal-based ones rely on storing exemplars of old classes to mitigate catastrophic forgetting, which limits real-world applications considering memory resources and privacy issues. In this paper, we propose a novel rehearsal-free CIL approach that learns continually via the synergy between two Complementary Learning Subnetworks. Our approach involves jointly optimizing a plastic CNN feature extractor and an analytical feed-forward classifier. The inaccessibility of historical data is tackled by holistically controlling the parameters of a well-trained model, ensuring that the decision boundary learned fits new classes while retaining recognition of previously learned classes. Specifically, the trainable CNN feature extractor provides task-dependent knowledge separately without interference;and the final classifier integrates task-specific knowledge incrementally for decision-making without forgetting. In each CIL session, it accommodates new tasks by attaching a tiny set of declarative parameters to its backbone, in which only one matrix per task or one vector per class is kept for knowledge retention. Extensive experiments on a variety of task sequences show that our method achieves competitive results against state-of-the-art methods, especially in accuracy gain, memory cost, training efficiency, and task-order robustness. Furthermore, to make the non-growing backbone (i.e., a model with limited network capacity) suffice to train on more incoming tasks, a graceful forgetting implementation on previously learned trivial tasks is empirically investigated. Copyright © 2023, The Authors. All rights reserved.

关键词： Decision making

来源：评论

学校读者我要写书评

暂无评论

Cluster Consensus Problem on Time-varying Higher-Order Interaction Networks

Cluster Consensus Problem on Time-varying Higher-Order Inter...

引用

IEEE Conference on Decision and control

作者： Haoyu Wei Lulu Pan Haibin Shao Dewei Li Wenbin Yu Department of Automation Shanghai Jiao Tong University and Key Laboratory of System Control and Information Processing Ministry of Education of China and Shanghai Engineering Research Center of Intelligent Control and Management Shanghai China

ISBN: (数字)9798350316339

ISBN: (纸本)9798350316346

This paper examines the cluster consensus problem of higher-order interaction networks on switching networks. The matrix-valued inter-agent coupling mechanism is employed to capture the higher-order interactions amongst neighboring agents. Specifically, for a higher-order interaction network with both positive definite and positive semi-definite coupling matrices, we examine the scenario in which the nullspace of positive semi-definite coupling matrices is constructed by a set of orthogonal basis vectors. From the perspective of the lower-order component graph -the projection of network structure onto the direction of each orthogonal basis–results on the explicit characterization of cluster consensus of higher-order interaction networks are provided. Moreover, for time-varying higher-order interaction networks, necessary and sufficient conditions are provided to ensure that a time-varying network shares the same steady-state values as the integral time-invariant network. Simulation results are provided to demonstrate the theory.

关键词： Couplings Sufficient conditions Network topology Simulation Switches Vectors Topology Steady-state Matrix decomposition Convergence

来源：评论

学校读者我要写书评

暂无评论

A2J-Transformer: Anchor-to-Joint Transformer Network for 3D Interacting Hand Pose Estimation from a Single RGB image

arXiv

引用

arXiv 2023年

作者： Jiang, Changlong Xiao, Yang Wu, Cunlin Zhang, Mingyang Zheng, Jinghong Cao, Zhiguo Zhou, Joey Tianyi Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan430074 China Alibaba Group China Singapore Singapore

3D interacting hand pose estimation from a single RGB image is a challenging task, due to serious self-occlusion and inter-occlusion towards hands, confusing similar appearance patterns between 2 hands, ill-posed joint position mapping from 2D to 3D, etc.. To address these, we propose to extend A2J-the state-of-the-art depth-based 3D single hand pose estimation method-to RGB domain under interacting hand condition. Our key idea is to equip A2J with strong local-global aware ability to well capture interacting hands’ local fine details and global articulated clues among joints jointly. To this end, A2J is evolved under Transformer’s non-local encoding-decoding framework to build A2J-Transformer. It holds 3 main advantages over A2J. First, self-attention across local anchor points is built to make them global spatial context aware to better capture joints’ articulation clues for resisting occlusion. Secondly, each anchor point is regarded as learnable query with adaptive feature learning for facilitating pattern fitting capacity, instead of having the same local representation with the others. Last but not least, anchor point locates in 3D space instead of 2D as in A2J, to leverage 3D pose prediction. Experiments on challenging InterHand 2.6M demonstrate that, A2J-Transformer can achieve state-of-the-art model-free performance (3.38mm MPJPE advancement in 2-hand case) and can also be applied to depth domain with strong generalization. The code is avaliable at https://***/ChanglongJiangGit/A2J-Transformer. Copyright © 2023, The Authors. All rights reserved.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Boosting Explore-Exploit Behavior for Navigating Vehicle by Maintaining Informative Topological Frontier

IEEE Transactions on Intelligent Vehicles

引用

IEEE Transactions on intelligent Vehicles 2024年 1-11页

作者： Liu, Qiming Cui, Xinru Liu, Zhe Wang, Hesheng Department of Automation Shanghai Jiao Tong University Shanghai China MoE Key Lab of Artificial Intelligence AI Institute Shanghai Jiao Tong University Shanghai China Department of Automation Key Laboratory of System Control and Information Processing of Ministry of Education Key Laboratory of Marine Intelligent Equipment and System of Ministry of Education Shanghai Engineering Research Center of Intelligent Control and Management Shanghai Jiao Tong University Shanghai China

Target search in unknown environments places high demands not only on an autonomous vehicle's ability to perceive and interpret target cues, but also on its conscious of collecting these cues by active exploration. While existing navigation methods have successfully built target-driven policies by maintaining memory of explored areas, there has been a lack of focus on facilitating target-aware exploration-the informative frontier information at unexplored yet visible areas is often overlooked. In this paper, we introduce a novel topology-based memory structure, Frontier-enhanced Topological Memory (FTM), and a Hierarchical Topology Encoding and Extraction (HTEE) module, fostering the autonomous vehicle's awareness of both environmental exploration and target approach. Specifically, FTM innovatively incorporates informative ghost nodes on traditional topological map to represent unexplored yet visible regions. We leverage an online-trained implicit scene representation to estimate the positions and generate features of these ghost nodes. The HTEE then employs implicit graph convolutions and attention mechanisms to extract cognitive information from FTM, taking into account the hierarchical memory structure, target cues, and current state. Our design bolsters cognitive navigation decisions. The experiments in the high-fidelity environments, including performance tests, visualizations, and interpretability experiments, validate the effectiveness of our approach in enhancing the vehicle's exploratory behavior. The improved exploration awareness for target cue collection, in turn, enhances the success rate and path efficiency of target search. Furthermore, we demonstrate the adaptability of our algorithm in real-world physical environments. IEEE

关键词： Data mining

来源：评论

学校读者我要写书评

暂无评论

Switched High-Gain Based Interval Observer Design for Harmonic Oscillator System 42

Switched High-Gain Based Interval Observer Design for Harmon...

引用

42nd Chinese control Conference, CCC 2023

作者： Ming, Tianhang Wang, Xiaoling Liu, Bo Su, Housheng College of Automation & Ai Nanjing University of Posts and Telecommunications Nanjing210023 China Nanjing210023 China College of Information Engineering Minzu University of China Beijing100081 China Huazhong University of Science and Technology Key Laboratory of Image Processing and Intelligence Control of Ministry of Education Wuhan430074 China

ISBN: (纸本)9789887581543

In this paper, we revisit the interval observer design problem of harmonic oscillator system with unknown but bounded external disturbance and measurement noise. With the bounding information of the external disturbance and measurement noise, a high-gain interval observer is firstly defined and constructed for the harmonic oscillator system to realize the interval estimation on the state and the robust control of the system. Then, a switched high-gain interval observer is proposed to reduce the influence of the measurement noise for a tighter interval width on the state estimation. Finally, some numerical simulations are given to verify the theoretical results. © 2023 Technical Committee on control Theory, Chinese Association of Automation.

关键词： Robust control

来源：评论

学校读者我要写书评

暂无评论

MPIB: An MPI-Based Bokeh Rendering Framework for Realistic Partial Occlusion Effects

arXiv

引用

arXiv 2022年

作者： Peng, Juewen Zhang, Jianming Luo, Xianrui Lu, Hao Xian, Ke Cao, Zhiguo Key Laboratory of Image Processing and Intelligent Control Ministry of Education School of AIA Huazhong University of Science and Technology China Adobe Research

Partial occlusion effects are a phenomenon that blurry objects near a camera are semi-transparent, resulting in partial appearance of occluded background. However, it is challenging for existing bokeh rendering methods to simulate realistic partial occlusion effects due to the missing information of the occluded area in an all-in-focus image. Inspired by the learnable 3D scene representation, Multiplane image (MPI), we attempt to address the partial occlusion by introducing a novel MPI-based high-resolution bokeh rendering framework, termed MPIB. To this end, we first present an analysis on how to apply the MPI representation to bokeh rendering. Based on this analysis, we propose an MPI representation module combined with a background inpainting module to implement high-resolution scene representation. This representation can then be reused to render various bokeh effects according to the controlling parameters. To train and test our model, we also design a ray-tracing-based bokeh generator for data generation. Extensive experiments on synthesized and real-world images validate the effectiveness and flexibility of this framework. Copyright © 2022, The Authors. All rights reserved.

关键词： Rendering (computer graphics)

来源：评论

学校读者我要写书评

暂无评论

A Detection Method Toward Yellow Dragon Disease Spreads Using Spark and Deep Learning 3

A Detection Method Toward Yellow Dragon Disease Spreads Usin...

引用

3rd International Conference on High Performance Computing and Communication Engineering, HPCCE 2023

作者： Xie, Wu Su, Zhenzhao Zhang, Huimin Kang, Ping Qin, Kun Fan, Yong Zhao, Quanyou Guangxi Key Laboratory of Image and Graphic Intelligent Processing Guilin University of Electronic Technology Guilin541214 China Key Laboratory of Education Blockchain and Intelligent Technology Ministry of Education Guangxi Normal University Guilin541004 China Lingchuan Science and Technology Information Institute Guilin541299 China Guilin Hivision Technology Co. Ltd. Guilin541004 China School of Mechanical and Electrical Engineering Guilin University of Electronic Technology Guilin541214 China School of Computer Science and Engineering Guilin University of Aerospace Technology Guilin541004 China

ISBN: (数字)9781510674653

ISBN: (纸本)9781510674646

The detection methods of yellow dragon disease spread via wood lice transmission networks like social networks are very important for diverse citrus trees and farmers. Although current methods have some detection accuracy or low cost, the detection processes are relatively troublesome and the detection cycles are long, making it be a difficult problem to apply them in large-scale orange farms as practical scenarios to detect citrus yellow dragon disease in a timely manner. A new method toward detecting citrus yellow dragon disease spread utilizing Spark and deep learning is proposed for this problem. By obtaining citrus field video stream data through high-definition cameras, and transferring the stream data to the Spark cluster through Kafka like intelligent agents, it is practicable to use the structured streaming component under the Spark framework via big data ecosphere to process video or image stream data transmitted via monitoring. We construct a citrus yellow dragon disease detection model via YOLOv7, and use self-made citrus yellow dragon disease images as training and testing data sets. The preliminary experimental results show that the new methods achieved an accuracy of 83.14%. To reduce the occurrence of missed and false detections, the shallow detection heads are added to the feature fusion networks for improvement, extracting and fusing shallow network information to try to improve the detection effects of yellow dragon disease. Then replace the convolution operation in the ELAN (Effective Long-range Aggregation Network) module with deep separable convolution to reduce the number of model parameters. The preliminary experimental results show that compared to the original YOLOv7 model, our improved citrus yellow dragon disease detection model with YOLOv7 has an accuracy improvement of 2.43%, maintaining a higher detection accuracy with lower time than before. © 2024 SPIE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

Velocity-Varying Target Tracking of Mobile Sensor Network Based on Flocking control

引用

Journal of Shanghai Jiaotong university(Science) 2021年第4期26卷 446-453页

作者： ZHANG Lalu Xiangxiang Dong YAO Lixiu CAI Yunze Key Laboratory of Marine Intelligent Equipment and System of Ministry of Education Shanghai Jiao Tong UniversityShanghai200240China Department of Automation Shanghai Jiao Tong UniversityShanghai200240China Key Laboratory of System Control and Information Processing of Ministry of Education Shanghai Jiao Tong UniversityShanghai200240China

Existing coupled distributed estimation and motion control strategies of mobile sensor networks present limitations in velocity-varying target tracking. Therefore, a velocity-varying target tracking algorithm based on flocking control is proposed herein. The Kalman-consensus filter is utilized to estimate the position, velocity and acceleration of a target. The flocking control algorithm with a velocity-varying virtual leader enables the position of the center of the mobile sensor network to converge to that of the target. By applying an effective cascading Lyapunov method, stability analysis is performed. Simulation results are provided to validate the feasibility of the proposed algorithm.

关键词： mobile sensor network velocity-varying target tracking flocking control Kalman-consensus filtering

来源：评论

学校读者我要写书评

暂无评论

Contour Extraction via Contextual Modulation

引用

Chinese Journal of Electronics 2023年第2期18卷 329-331页

作者： Wentao Huang Licheng Jiao Yuelei Xu Ye Sun Jianhua Jia Key Laboratory of Intelligent Perception and Image Understanding of Ministry of Education Institute of Intelligent Information Processing Xidian University Xi'an China Department of Avionics Engineering Engineering College of Air Force Engineering University Xi'an China

Contour extraction is an important task in image processing and computer vision. The contextual modulation is a universal phenomenon in the primary visual cortex (V1). A biologically motivated computational model is presented for contour extraction in this paper. Two mechanisms of contextual modulation, surround suppression and collinear facilitation, are integrated in this model. We obtain good results via this model to extract contours from images with noise and texture backgrounds. This work provides a biologically motivated approach with great potential for computer vision.

关键词： Visualization Computer vision Computational modeling Biological system modeling image processing Modulation Organizations

来源：评论

学校读者我要写书评

暂无评论

A Probabilistic Topology Inference Method for Networked Dynamical System via Single Excitation

A Probabilistic Topology Inference Method for Networked Dyna...

引用

IEEE Conference on Decision and control

作者： Qing Jiao Yushan Li Jianping He Dept. of Automation Key Laboratory of System Control and Information Processing Ministry of Education of China and Shanghai Engineering Research Center of Intelligent Control and Management Shanghai Jiao Tong University Shanghai China

ISBN: (数字)9798350316339

ISBN: (纸本)9798350316346

A growing number of works have investigated inferring the topology of networked dynamical systems from observations, such as to better understand the system behaviour. Despite the tremendous advances, most of them require the observations to be abundant. This paper focuses on inferring the topology by injecting single excitation on a node and collecting several steps of noisy observations. The problem is challenging because the noises cannot be depressed in several observations and are mixed with the injected excitation, making it hard to directly reveal the topology. To practice, we develop a probabilistic method based on the hypothesis test framework. First, we infer the neighbors that are within h-hop of the excited node and derive the accuracy guarantees. Then, we extend the method to infer the exact h-hop neighbors. A computable lower bound for the accuracy probability is established to provide confidence support in the inference procedures. Furthermore, we give the conditions of excitation input to ensure a desired inference probability, which provides guidance for the input design. Numerical simulations are conducted to verify the effectiveness of the proposed method.

关键词： Lower bound Accuracy Network topology Noise Probabilistic logic Numerical simulation Topology Noise measurement Reliability Dynamical systems

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：