检索结果-内蒙古大学图书馆

Enhancing Environmental Robustness in Few-shot Learning via Conditional Representation Learning

IEEE Transactions on Image Processing 2025年 34卷 3489-3502页

作者： Guo, Qianyu Wu, Jingrong Wu, Tianxing Wang, Haofen Ge, Weifeng Zhang, Wenqiang Fudan University School of Computer Science China Shanghai Jiao Tong University School of Medicine Shanghai Institute of Virology China Southeast University School of Computer Science and Engineering Nanjing 210096 China Tongji University College of Design and Innovation China Fudan University Engineering Research Center of AI and Robotics Ministry of Education Academy for Engineering and Technology Shanghai 20043 China

Few-shot learning (FSL) has recently been extensively utilized to overcome the scarcity of training data in domain-specific visual recognition. In real-world scenarios, environmental factors such as complex backgrounds, varying lighting conditions, long-distance shooting, and moving targets often cause test images to exhibit numerous incomplete targets or noise disruptions. However, current research on evaluation datasets and methodologies has largely ignored the concept of “environmental robustness”, which refers to maintaining consistent performance in complex and diverse physical environments. This neglect has led to a notable decline in the performance of FSL models during practical testing compared to their training performance. To bridge this gap, we introduce a new real-world multi-domain few-shot learning (RD-FSL) benchmark, which includes four domains and six evaluation datasets. The test images in this benchmark feature various challenging elements, such as camouflaged objects, small targets, and blurriness. Our evaluation experiments reveal that existing methods struggle to utilize training images effectively to generate accurate feature representations for challenging test images. To address this problem, we propose a novel conditional representation learning network (CRLNet) that integrates the interactions between training and testing images as conditional information in their respective representation processes. The main goal is to reduce intra-class variance or enhance inter-class variance at the feature representation level. Finally, comparative experiments reveal that CRLNet surpasses the current state-of-the-art methods, achieving performance improvements ranging from 6.83% to 16.98% across diverse settings and backbones. © 1992-2012 IEEE.

关键词： Conditional Representation Learning Environmental Robustness Few-Shot Learning Visual Recognition

来源：评论

学校读者我要写书评

暂无评论

From instantaneous schedulability to worst case schedulability: a significant moment approach

引用

Cyber-Physical Systems 2024年

作者： Yao, Ningshi Zhang, Fumin Department of Electrical and Computer Engineering George Mason University FairfaxVA United States Cheng Kar-Shun Robotics Institute Hong Kong University of Science and Technology Hong Kong Hong Kong SAR Hong Kong

The method of significant moment analysis has been used to derive instantaneous schedulability tests for real-time tasks. However, the instantaneous schedulability can only be checked within a finite time window. On the other hand, worst-case schedulability guarantees schedulability of tasks for infinite time. This paper derives the classical worst-case schedulability conditions for pre-emptive periodic tasks starting from instantaneous schedulability, hence unifying the two notions of schedulability. The results provide a rigorous justification for the critical time instants being the worst case for scheduling pre-emptive periodic tasks. The paper also shows that the critical time instant is not the only worst-case moments. © 2024 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.

关键词： critical instant analysis instantaneous schedulability significant moment analysis Worst-case schedulability

来源：评论

学校读者我要写书评

暂无评论

Pearl Millet Disease Detection with Transfer Learning and cGAN-Driven Data Augmentation

Pearl Millet Disease Detection with Transfer Learning and cG...

引用

Pervasive Computing and Social Networking (ICPCSN), International Conference on

作者： Melvin A S X. Anitha Mary Kumudha Raimond I. Johnson Karunya Institute of Technology and Sciences Coimbatore India Department of Robotics Karunya Institute of Technology and Sciences Coimbatore India Division of Computer Science and Engineering Karunya Institute of Technology and Sciences Coimbatore India Department of Plant Pathology TamilNadu Agricultural University Coimbatore India

ISBN: (数字)9798331535193

ISBN: (纸本)9798331535209

In this study, an advanced method to detect disease in Pearl Millet is proposed wherein Transfer Learning along with Conditional Generative Adversarial Networks (cGANs) are applied for data augmentation. For addressing dataset quality and imbalance problem, the work uses cGANs to generate realistic synthetic images of diseased millet leaves with latent dimensions of 50, 100, and 500. This augmented dataset of plant roots helps improve the classification performance as it provides more quantity and more diversity of training data. For the classification task, six fine-tuned models are used; these are: ResNet-50, ResNet-101, ResNet-152, InceptionResNet, InceptionV3, and VGG-19; optimizer is Stochastic Gradient Descent (SGD). The performance of these models was trained by 100 epochs and evaluated by accuracy, precision, recall, F1 score and ROC curve. However, ResNet-50 and ResNet-152 are among those and their remarkable accuracy of 99% is better than the previous work’s 98% accuracy, proving this integrated approach to disease detection can be effective. The results show that transferring knowledge from a prior set of training data, by applying transfer learning, improves disease classification accuracy, along with applying data augmentation, which allows the creation of additional data, and robust model evaluation, to ensure accurate evaluation.

关键词： Pervasive computing Accuracy Social networking (online) Transfer learning Training data Stochastic processes Data augmentation Generative adversarial networks Data models Diseases

来源：评论

学校读者我要写书评

暂无评论

Absolute Pose from One or Two Scaled and Oriented Features

Absolute Pose from One or Two Scaled and Oriented Features

引用

Conference on computer Vision and Pattern Recognition (CVPR)

作者： Jonathan Ventura Zuzana Kukelova Torsten Sattler Dániel Baráth Department of Computer Science & Software Engineering Cal Poly San Luis Obispo Faculty of Electrical Engineering Visual Recognition Group Czech Technical University in Prague Czech Institute of Informatics Robotics and Cybernetics Czech Technical University in Prague Department of Computer Science Computer Vision and Geometry Group ETH Zürich

ISBN: (数字)9798350353006

ISBN: (纸本)9798350353013

Keypoints used for image matching often include an estimate of the feature scale and orientation. While recent work has demonstrated the advantages of using feature scales and orientations for relative pose estimation, relatively little work has considered their use for absolute pose estimation. We introduce minimal solutions for absolute pose from two oriented feature correspondences in the general case, or one scaled and oriented correspondence given a known vertical direction. Nowadays, assuming a known direction is not particularly restrictive as modern consumer devices, such as smartphones or drones, are equipped with Inertial Measurement Units (IMU) that provide the gravity direction by default. Compared to traditional absolute pose methods requiring three point correspondences, our solvers need a smaller minimal sample, reducing the cost and complexity of robust estimation. Evaluations on large-scale and public real datasets demonstrate the advantage of our methods for fast and accurate localization in challenging conditions. Code is available at https://***/danini/absolute-pose-from-oriented-and-sealed-features.

关键词： Location awareness Accuracy Simultaneous localization and mapping Measurement units Pose estimation Noise Inertial navigation

来源：评论

学校读者我要写书评

暂无评论

Temporally Robust Multi-Agent STL Motion Planning in Continuous Time

arXiv

引用

arXiv 2023年

作者： Verhagen, Joris Lindemann, Lars Tumova, Jana The Division of Robotics Perception and Learning School of Electrical Engineering and Computer Science Kth Royal Institute of Technology Stockholm Sweden The Thomas Lord Department of Computer Science University of Southern California Los AngelesCA United States

Signal Temporal Logic (STL) is a formal language over continuous-time signals (such as trajectories of a multiagent system) that allows for the specification of complex spatial and temporal system requirements (such as staying sufficiently close to each other within certain time intervals). To promote robustness in multi-agent motion planning with such complex requirements, we consider motion planning with the goal of maximizing the temporal robustness of their joint STL specification, i.e. maximizing the permissible time shifts of each agent's trajectory while still satisfying the STL specification. Previous methods presented temporally robust motion planning and control in a discrete-time Mixed Integer Linear Programming (MILP) optimization scheme. In contrast, we parameterize the trajectory by continuous Bézier curves, where the curvature and the time-traversal of the trajectory are parameterized individually. We show an algorithm generating continuous-time temporally robust trajectories and prove soundness of our approach. Moreover, we empirically show that our parametrization realizes this with a considerable speed-up compared to state-of-the-art methods based on constant interval time discretization. Copyright © 2023, The Authors. All rights reserved.

关键词： Trajectories

来源：评论

学校读者我要写书评

暂无评论

Domain-Invariant Similarity Activation Map Contrastive Learning for Retrieval-Based Long-Term Visual Localization

引用

IEEE/CAA Journal of Automatica Sinica 2022年第2期9卷 313-328页

作者： Hanjiang Hu Hesheng Wang Zhe Liu Weidong Chen Department of Automation Shanghai Jiao Tong UniversityShanghai 200240China IEEE Department of Automation Institute of Medical RoboticsKey Laboratory of System Control and Information Processing of Ministry of EducationKey Laboratory of Marine Intelligent Equipment and System of Ministry of EducationShanghai Jiao Tong UniversityShanghai 200240China Department of Computer Science and Technology University of CambridgeCambridge CB30FDUnited Kingdom

Visual localization is a crucial component in the application of mobile robot and autonomous *** retrieval is an efficient and effective technique in image-based localization *** to the drastic variability of environmental conditions,e.g.,illumination changes,retrievalbased visual localization is severely affected and becomes a challenging *** this work,a general architecture is first formulated probabilistically to extract domain-invariant features through multi-domain image ***,a novel gradientweighted similarity activation mapping loss(Grad-SAM)is incorporated for finer localization with high *** also propose a new adaptive triplet loss to boost the contrastive learning of the embedding in a self-supervised *** final coarse-to-fine image retrieval pipeline is implemented as the sequential combination of models with and without Grad-SAM *** experiments have been conducted to validate the effectiveness of the proposed approach on the CMU-Seasons *** strong generalization ability of our approach is verified with the RobotCar dataset using models pre-trained on urban parts of the CMU-Seasons *** performance is on par with or even outperforms the state-of-the-art image-based localization baselines in medium or high precision,especially under challenging environments with illumination variance,vegetation,and night-time ***,real-site experiments have been conducted to validate the efficiency and effectiveness of the coarse-to-fine strategy for localization.

关键词： Deep representation learning place recognition visual localization

来源：评论

学校读者我要写书评

暂无评论

Exploring Intermittent Dynamics of Neural Activity in a Single-Pair System of Excitatory and Inhibitory Neurons

Exploring Intermittent Dynamics of Neural Activity in a Sing...

引用

Emerging Techniques in Computational Intelligence (ICETCI), International Conference on

作者： Akio Sugawara Sou Nobukawa Nobuhiko Wagatsuma Keiichiro Inagaki Department of Computer Science Chiba Institute of Technology Narashino Japan Department of Information Science Toho University Funabashi Japan Department of Artificial Intelligence and Robotics Chubu University Kasugai Japan

A long-tailed property has been observed at various levels of the brain, ranging from neural population activity to the level of cognitive neurodynamics. Specifically, at the cognitive neurodynamics level, a phenomenon called perceptual alternation displays perceptual durations that follow a gamma or log-normal distribution. Moreover, this perceptual alternation becomes nondeterministic. The occurrence of alternation phenomena has also been noticed in neural populations at the cognitive neurodynamics level. Even in a system consisting of a single pair of excitatory and inhibitory neurons, referred to as chaos-chaos intermittency (CCI), a similar intermittent alternation of neural activity emerges, involving intermittent transitions between multiple isolated attractors. In this study, our hypothesis was that nondeterminism and long-tailed properties can emerge in this single-pair system. To test this hypothesis, we evaluated the determinism of two types of dynamics in the system that couples excitatory and inhibitory neurons: 1) transitions between attractors and 2) behavior within attractors. This evaluation was performed on multiple time scales using multi-scale entropy analysis (MSE) and an iterated amplitude-adjusted Fourier transform (IAAFT) surrogate. The results demonstrated that the transitions between attractors were nondeterministic, while the behavior within attractors was deterministic. Furthermore, a long-tailed property manifested in the duration of the transitions between attractors, possibly resulting from long-term evaluation in the chaotic unstable orbit. The emergence of this long-tailed property was attributed to the low frequency CCI around the attractor-merging bifurcation. This discovery contributes to the understanding of the long-tailed properties exhibited by neural activity across multiple levels.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Parallel and Asynchronous Blob Shape Descriptor Implemented in CMOS Technology

A Parallel and Asynchronous Blob Shape Descriptor Implemente...

引用

International Conference on Microelectronics, MIEL

作者： P. Bogacki M. Długosz T. Talaśka R. Długosz Aptiv Services Poland Kraków Poland Institute of Telecommunications Faculty of Computer Science Electronics and Telecommunications AGH University of Science and Technology Kraków Poland Faculty of Control Robotics and Electrical Engineering Institute of Automation and Robotics Division of Signal Processing and Electronic Systems Poznan University of Technology Poznan Poland Faculty of Telecommunication Computer Science and Electrical Engineering Bydgoszcz University of Science and Technology Bydgoszcz Poland

The paper presents a family of novel light blob shape descriptors for use in selected active safety algorithms used in Advanced Driver Assistance Systems (ADAS). One of the motivations was to obtain a descriptor that would ensure low computational complexity. This makes it easy to implement both in software and hardware. One assumption is that the location of the center of a given light spot is approximately known. The principle of its operation is then to count white pixels in selected directions, starting from this central point. A key issue here is an efficient way of determining indexes of particular pixels belonging to the image patch, as well as the location of points representing places where the white area turns into black. In the case of a hardware implementation, this can be done using a parallel circuit operating in asynchronous mode, without the need for a control clock.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A study on the noise reduction method in turbid water environments using Spatial frequency characteristics

A study on the noise reduction method in turbid water enviro...

引用

International Symposium on Environmentally Conscious Design and Inverse Manufacturing (EcoDesign)

作者： Ayumu Koura Hyun-Woo Kim Myungjin Cho Min-Chul Lee Graduate School of Computer Science and Systems Engineering Kyushu Institute of Technology Iizuka city Japan School of ICT Robotics and Mechanical Engineering IITC Hankyong National University Anseong-si Korea

ISBN: (数字)9798350363043

ISBN: (纸本)9798350363050

Scattering noise reduction is a challenging project to remove noise under scattering media conditions such as fog or turbid water. In previous study, blurring caused by scattering medium particles in fog or turbid water was removed by low-pass filtering. However, in this method, the accurate color information of the object may be lost due to rough low-frequency filtering. In addition, it is difficult to the obtain object shape information in heavy turbid environments. To solve this problem, we propose a new de-scattering algorithm that combines Fourier-domain filtering and integral imaging by creating a filter tailored to the frequency to be removed. To prove the proposed method, we perform a numerical analysis of experimental results using conventional and proposed methods.

关键词： Water Three-dimensional displays Image color analysis Shape Numerical analysis Noise reduction Scattering Reconstruction algorithms Colored noise Image reconstruction

来源：评论

学校读者我要写书评

暂无评论

Lightweight 2D Human Pose Estimation for Fitness Coaching System 36

Lightweight 2D Human Pose Estimation for Fitness Coaching Sy...

引用

36th International Technical Conference on Circuits/Systems, computers and Communications, ITC-CSCC 2021

作者： Jeon, Hobeom Yoon, Youngwoo Kim, Dohyung University of Science and Technology Department of Computer Software Korea Republic of Electronics and Telecommunications Research Institute Intelligent Robotics Research Division Korea Republic of

ISBN: (纸本)9781665435536

In this paper, we describe lightweight 2D human pose estimation for a fitness coaching system. To achieve real-time inference speed on mobile devices, we propose the online pose distillation learning strategy that trains large teacher networks and small student networks simultaneously. The proposed lightweight model requires significantly lower computation as well as shows competitive pose estimation accuracy on the COCO keypoint detection dataset. In addition, the model was fine-tuned on a fitness dataset for the fitness coaching application where various fitness postures were accessed. © 2021 IEEE.

关键词： Distillation

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：