检索结果-内蒙古大学图书馆

ORA3D: Overlap Region Aware Multi-view 3D Object Detection 33

学校读者我要写书评

暂无评论

ORA3D: Overlap Region Aware Multi-view 3D Object Detection

33rd British Machine Vision Conference Proceedings, BMVC 2022

作者： Roh, Wonseok Chang, Gyusam Moon, Seokha Nam, Giljoo Kim, Chanyoung Kim, Younghyun Kim, Sangpil Kim, Jinkyu Department of Artificial Intelligence Korea University Seoul Korea Republic of School of Computing KAIST Daejeon Korea Republic of Department of Computer Science and Engineering Korea University Seoul Korea Republic of Autonomous Driving Center Hyundai Motor Company R&D Division Seoul Korea Republic of

Current multi-view 3D object detection methods often fail to detect objects in the overlap region properly, and the networks' understanding of the scene is often limited to that of a monocular detection network. Moreover, objects in the overlap region are often largely occluded or suffer from deformation due to camera distortion, causing a domain shift. To mitigate this issue, we propose using the following two main modules: (1) Stereo Disparity Estimation for Weak Depth Supervision and (2) Adversarial Overlap Region Discriminator. The former utilizes the traditional stereo disparity estimation method to obtain reliable disparity information from the overlap region. Given the disparity estimates as supervision, we propose regularizing the network to fully utilize the geometric potential of binocular images and improve the overall detection accuracy accordingly. Further, the latter module minimizes the representational gap between non-overlap and overlapping regions. We demonstrate the effectiveness of the proposed method with the nuScenes large-scale multi-view 3D object detection data. Our experiments show that our proposed method outperforms current state-of-the-art models, i.e., DETR3D and BEVDet. © 2022. The copyright of this document resides with its authors. It may be distributed unchanged freely in print or electronic forms.

关键词： Object detection

Multimodal Knowledge Graph Vision-Language Models for Precision Retrieval and Inference in Embedded Systems

学校读者我要写书评

暂无评论

Multimodal Knowledge Graph Vision-Language Models for Precis...

2024 IEEE International Conference on Robotics and Biomimetics, ROBIO 2024

作者： Li, Mingyi Li, Jiayin Liu, Hui Han, Lijin Liao, Dengting Zhang, Zekun Gao, Zong Liu, Baoshuai Jin, Zhongfeng School of Mechanical Engineering Beijing Institute of Technology Beijing100081 China School of Computing and Data Science Xiamen University Malaysia Selangor 43900 Malaysia Shanghai Zhangjiang Institute of Mathematicas Zhangjiang Institute of Shanghai Shanghai201203 China China North Vehicle Research Institute General Technology Department 100072 China Coordination Center of China GNational Computer Network Emergency Response Technical Team 100083 China

ISBN: (纸本)9781665481090

In light of the challenges associated with enhancing multimodal knowledge graph retrieval and improving embodied intelligence in dynamic environments, we propose the Multimodal Knowledge Graph Vision-Language System (MKGVL). This system is designed to enhance the reasoning and feedback capabilities of embodied intelligent systems by continuously updating the knowledge graph through real-time feedback from the Vision-Language Model (VLM). By integrating visual encoders, language models, and knowledge graph networks, MKGVL constructs a unified multimodal representation that enables adaptive decision-making and responsiveness to environmental changes. Experimental results show that MKGVL outperforms existing models in fine-grained retrieval tasks, achieving an 11.5% improvement in Rank1 accuracy and a mean Average Precision (mAP) of 97.49%. Further evaluations conducted on datasets such as ARKitScenes, MultiScan, and 3RScan highlight the model's robustness and adaptability. Additionally, MKGVL's deployment on embedded platforms like Jetson Orin demonstrates its efficiency in real-time multimodal tasks, particularly in resource-constrained environments. These findings underscore MKGVL's ability to deliver accurate and efficient multimodal processing, making it a strong solution for adaptive knowledge-based systems in complex settings. © 2024 IEEE.

关键词： Knowledge graph

Communicative Robot Signals: Presenting a New Typology for Human-Robot Interaction

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Holthaus, Patrick Schulz, Trenton Lakatos, Gabriella Soma, Rebekka University of Hertfordshire School of Physics Engineering and Computer Science Hatfield United Kingdom Norwegian Computing Center Department of Applied Research in ICT Oslo Norway University of Oslo Department of Informatics Oslo Norway

We present a new typology for classifying signals from robots when they communicate with humans. For inspiration, we use ethology, the study of animal behaviour and previous efforts from literature as guides in defining the typology. The typology is based on communicative signals that consist of five properties: the origin where the signal comes from, the deliberateness of the signal, the signal’s reference, the genuineness of the signal, and its clarity (i.e., how implicit or explicit it is). Using the accompanying worksheet, the typology is straightforward to use to examine communicative signals from previous human-robot interactions and provides guidance for designers to use the typology when designing new robot behaviours. © 2023, CC BY.

关键词： Human robot interaction

AgileWatts: An Energy-Efficient CPU Core Idle-State Architecture for Latency-Sensitive Server Applications 22

学校读者我要写书评

暂无评论

AgileWatts: An Energy-Efficient CPU Core Idle-State Architec...

Proceedings of the 55th Annual IEEE/ACM International Symposium on Microarchitecture

作者： Jawad Haj Yahya Haris Volos Davide B. Bartolini Georgia Antoniou Jeremie S. Kim Zhe Wang Kleovoulos Kalaitzidis Tom Rollet Zhirui Chen Ye Geng Onur Mutlu Yiannakis Sazeides Rivos Inc Mountain View California United States Computer Architecture Laboratory Department of Computer Science University of Cyprus Computing Systems Lab Huawei Zurich Research Center Department of Information Technology and Electrical Engineering ETH Zurich Switzerland

ISBN: (纸本)9781665462723

User-facing applications running in modern datacenters exhibit irregular request patterns and are implemented using a multitude of services with tight latency requirements (30--250μs). These characteristics render existing energy-conserving techniques ineffective when processors are idle due to the long transition time (order of 100μs) from a deep CPU core idle power state (C-state). While prior works propose management techniques to mitigate this inefficiency, we tackle it at its root with AgileWatts (AW): a new deep CPU core C-state architecture optimized for datacenter server processors targeting latency-sensitive *** drastically reduces the transition latency from deep CPU core idle power states while retaining most of their power savings based on three key ideas. First, AW eliminates the latency (several microseconds) of saving/restoring the core context when powering-off/-on the core in a deep idle state by i) implementing medium-grained power-gates, carefully distributed across the CPU core, and ii) retaining context in the power-ungated domain. Second, AW eliminates the flush latency (several tens of microseconds) of the L1/L2 caches when entering a deep idle state by keeping L1/L2 content power-ungated. A small control logic also remains ungated to serve cache coherence traffic. AW implements cache sleep-mode and leakage reduction for the power-ungated domain by lowering a core's voltage to the minimum operational level. Third, using a state-of-the-art power efficient all-digital phase-locked loop (ADPLL) clock generator, AW keeps the PLL active and locked during the idle state, cutting microseconds of wake-up latency at negligible power *** evaluation with an accurate industrial-grade simulator calibrated against an Intel Skylake server shows that AW reduces the energy consumption of Memcached by up to 71% (35% on average) with <1% end-to-end performance degradation. We observe similar trends for other evaluated services (MySQL and Kafka).

关键词：

Deep Learning for Camera Calibration and Beyond: A Survey

学校读者我要写书评

暂无评论

arXiv 2023年

作者： Liao, Kang Nie, Lang Huang, Shujuan Lin, Chunyu Zhang, Jing Zhao, Yao Gabbouj, Moncef Tao, Dacheng Institute of Information Science Beijing Jiaotong University Beijing100044 China Visual Intelligence + X International Joint Laboratory of the Ministry of Education China Department of Computing Sciences Tampere University Tampere33101 Finland School of Computer Science Faculty of Engineering The University of Sydney Australia

Camera calibration involves estimating camera parameters to infer geometric features from captured sequences, which is crucial for computer vision and robotics. However, conventional calibration is laborious and requires dedicated collection. Recent efforts show that learning-based solutions have the potential to be used in place of the repeatability works of manual calibrations. Among these solutions, various learning strategies, networks, geometric priors, and datasets have been investigated. In this paper, we provide a comprehensive survey of learning-based camera calibration techniques, by analyzing their strengths and limitations. Our main calibration categories include the standard pinhole camera model, distortion camera model, cross-view model, and cross-sensor model, following the research trend and extended applications. As there is no unified benchmark in this community, we collect a holistic calibration dataset that can serve as a public platform to evaluate the generalization of existing methods. It comprises both synthetic and real-world data, with images and videos captured by different cameras in diverse scenes. Toward the end of this paper, we discuss the challenges and provide further research directions. To our knowledge, this is the first survey for the learning-based camera calibration (spanned 10 years). The summarized methods, datasets, and benchmarks are available and will be regularly updated at https://***/KangLiao929/Awesome-Deep-Camera-Calibration. Copyright © 2023, The Authors. All rights reserved.

关键词： Pinhole cameras

A Hybrid Solution for Secure Privacy-Preserving Cloud Storage & Information Retrieval

学校读者我要写书评

暂无评论

Human-centric computing and Information sciences 2023年 13卷

作者： Kumar, Ankit Aljrees, Turki Hsieh, Sun-Yuan Singh, Kamred Udham Singh, Teekam Raja, Linesh Samriya, Jitendra Kumar Mundotiya, Rajesh Kumar Department of Computer Engineering & Applications GLA University UP Mathura India College of Computer Science and Engineering University of Hafr Al-Batin Eastern Province Saudi Arabia Department of Computer Science and Information Engineering Institute of Medical Information Institute of Manufacturing Information and Systems Center for Innovative FinTech Business Models and International Center for the Scientific Development of Shrimp Aquaculture National Cheng Kung University Tainan Taiwan School of Computing Graphic Era Hill University Dehradun India Department of Computer Science and Information Engineering National Cheng Kung University Tainan Taiwan Department of Computer Science and Engineering Graphic Era Deemed to be University Dehradun India Department of Computer Application Manipal University Jaipur Jaipur India Department of Computer Science and Engineering National Institute of Technology Delhi Delhi India School of Computer Science University of Petroleum and Energy Studies Dehradun India

Cloud storage is an emerging archetype used by businesses for data storage. Clients require easy access to data in the cloud, which helps clients with limited computing power move their high-value, high-risk principle jobs to the cloud. The primary concern of this study is towards flawless verifying, checking data bundles, and determining modules to carry out the project. As discussed at several research platforms, each client is concerned about data storage and retrieval. Data security is critical in cloud computing, and thus monitoring has evolved into a vigilance that abstracts the monitoring process. Request made by clients for security planning is always computationally exclusive. As a result, each customer is concerned as to their verified condition. This paper will highlight a secure cloud communication method for client data to address the afore-mentioned issues. Synchronization creates secret hare keys, combined with a model-encrypted data packet sent to the cloud. These secure data are then sent over a public network, subject to monitoring to avert data leakage to an unauthorized party. They lack the cohesion to outsource data storage to the cloud deliberately. As an outcome of this model's distribution, the perception of authorized data owners would be changed regarding cloud storage up to 97.6%, addressing the trust issues between data owners and cloud ace affiliations © 2023, Human-centric computing and Information *** Rights Reserved.

关键词： Cryptography

STRPM: A Spatiotemporal Residual Predictive Model for High-Resolution Video Prediction

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Chang, Zheng Zhang, Xinfeng Wang, Shanshe Ma, Siwei Gao, Wen School of Computer Science and Technology University of Chinese Academy of Sciences Beijing China Institute of Computing Technology Chinese Academy of Sciences Beijing China National Engineering Research Center of Visual Technology School of Computer Science Peking University Beijing China

Although many video prediction methods have obtained good performance in low-resolution (64∼128) videos, predictive models for high-resolution (512∼4K) videos have not been fully explored yet, which are more meaningful due to the increasing demand for high-quality videos. Compared with low-resolution videos, high-resolution videos contain richer appearance (spatial) information and more complex motion (temporal) information. In this paper, we propose a Spatiotemporal Residual Predictive Model (STRPM) for high-resolution video prediction. On the one hand, we propose a Spatiotemporal Encoding-Decoding Scheme to preserve more spatiotemporal information for high-resolution videos. In this way, the appearance details for each frame can be greatly preserved. On the other hand, we design a Residual Predictive Memory (RPM) which focuses on modeling the spatiotemporal residual features (STRF) between previous and future frames instead of the whole frame, which can greatly help capture the complex motion information in high-resolution videos. In addition, the proposed RPM can supervise the spatial encoder and temporal encoder to extract different features in the spatial domain and the temporal domain, respectively. Moreover, the proposed model is trained using generative adversarial networks (GANs) with a learned perceptual loss (LP-loss) to improve the perceptual quality of the predictions. Experimental results show that STRPM can generate more satisfactory results compared with various existing methods. © 2022, CC BY.

关键词： Signal encoding

TOWARDS NONLINEARITY. THE p-REGULARITY THEORY. APPLICATIONS AND DEVELOPMENTS.

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Bednarczuk, E. Brezhneva, O. Leśniewski, K. Prusińska, A. Tret’Yakov, A. Warsaw University of Technology Poland Department of Mathematics Miami University OxfordOH United States System Research Institute Polish Academy of Sciences Poland Faculty of Science University of Siedlce Poland Dorodnicyn Computing Center Federal Research Center "Computer Science and Control " Russian Academy of Sciences Moscow119333 Russia

We present recent advances in the analysis of nonlinear equations with singular operators and nonlinear optimization problems with constraints given by singular mappings. The results are obtained within the framework of p-regularity theory, which has developed successfully over the last forty years. We illustrate the theory with its applications to degenerate problems in various areas of mathematics. In particular, we address the problem of describing the tangent cone to the solution set of nonlinear equations in a singular case. The structure of p-factor operators is used to propose optimality conditions and construct numerical methods for solving degenerate nonlinear equations and optimization problems. The methods presented in the paper can be considered as the first numerical approaches targeting solutions of degenerate problems, such as the Van der Pol differential equation, boundary-value problems with a small parameter, partial differential equations where Poincaré’s method of small parameter fails, nonlinear degenerate dynamical systems, and others. There are various practical applications for the theory of p-regularity, including structural engineering, composite materials, and material design. For instance, the theory can be applied to analyze the behavior of materials with irregular or complex properties. By considering higher-order derivatives, it becomes possible to model and predict the response of materials to external forces, such as stress or temperature variations. In geophysics, the p-regularity theory can be utilized to analyze and interpret complex data obtained from seismic surveys, gravity measurements, or electromagnetic surveys. The theory also finds applications in the analysis of nonlinear differential equations arising in control systems, geometric and topological analysis, biomechanics, and many other fields. The theory offers a valuable approach to analyzing and understanding nonlinear phenomena in situations where classical regularity

关键词： Boundary value problems

Unsupervised Cortical Surface Registration Network for Aligning Gyralnet

学校读者我要写书评

暂无评论

Unsupervised Cortical Surface Registration Network for Align...

IEEE International Symposium on Biomedical Imaging

作者： Jiale Cheng Fenqiang Zhao Dan Hu Chao Cao Zhengwang Wu Xinrui Yuan Kangfu Han Lu Zhang Tianming Liu Dajiang Zhu Gang Li Department of Radiology and Biomedical Research Imaging Center University of North Carolina at Chapel Hill NC USA Joint Department of Biomedical Engineering University of North Carolina at Chapel Hill and North Carolina State University NC USA Department of Computer Science and Engineering University of Texas at Arlington TX USA Department of Computer Science Indiana University Indianapolis IN USA School of Computing University of Georgia GA USA

ISBN: (数字)9798331520526

ISBN: (纸本)9798331520533

The cortical 3-hinge gyrus (3HG) and its network (GyralNet) play key roles in understanding the regularity and variability of brain structure and function. However, existing cortical surface registration methods overlook these features, resulting in suboptimal alignment across subjects. Currently, no 3HG and GyralNet atlas exist for registration, and generation of the corresponding atlas requires extensive runtime using traditional methods. To enable better registration of these features, we introduce an unsupervised learning framework to jointly develop 3HGs and GyralNet atlas and register the individual cortical features onto the atlas. To incorporate the graph structure of 3HGs and GyralNet into the registration network, we convert them into surface distance maps, facilitating effective integration. To effectively learn large deformations, a multi-level spherical registration network based on spherical U-Net is introduced to perform registration in a coarse-to-fine manner. Experiments demonstrate our approach's ability to generate 3HGs and GyralNet atlas with detailed patterns and effectively improve registration accuracy.

关键词： Brain Runtime Accuracy Deformation Registers Unsupervised learning Biomedical imaging