检索结果-内蒙古大学图书馆

Adapting Human Mesh Recovery with Vision-Language Feedback

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Xu, Chongyang Huang, Buzhen Zhang, Chengfang Feng, Ziliang Wang, Yangang College of Computer Science Sichuan University Chengdu610065 China Key Laboratory of Measurement and Control of Complex Systems of Engineering Ministry of Education The School of Automation Southeast University Nanjing210096 China Intelligent Policing Key Laboratory of Sichuan Province Sichuan Police College Luzhou646000 China

Human mesh recovery can be approached using either regression-based or optimization-based methods. Regression models achieve high pose accuracy but struggle with model-to-image alignment due to the lack of explicit 2D-3D correspondences. In contrast, optimization-based methods align 3D models to 2D observations but are prone to local minima and depth ambiguity. In this work, we leverage large vision-language models (VLMs) to generate interactive body part descriptions, which serve as implicit constraints to enhance 3D perception and limit the optimization space. Specifically, we formulate monocular human mesh recovery as a distribution adaptation task by integrating both 2D observations and language descriptions. To bridge the gap between text and 3D pose signals, we first train a text encoder and a pose VQ-VAE, aligning texts to body poses in a shared latent space using contrastive learning. Subsequently, we employ a diffusion-based framework to refine the initial parameters guided by gradients derived from both 2D observations and text descriptions. Finally, the model can produce poses with accurate 3D perception and image consistency. Experimental results on multiple benchmarks validate its effectiveness. The code will be made publicly available. © 2025, CC BY.

关键词： Mesh generation

S2KAN-SLAM: Elastic Neural LiDAR SLAM with SDF Submaps and Kolmogorov-Arnold Networks

学校读者我要写书评

暂无评论

IEEE Transactions on Circuits and systems for Video Technology 2025年

作者： Wang, Zhong Zhang, Lin Wang, Hesheng Shanghai Jiao Tong University State Key Laboratory of Avionics Integration and Aviation System-of-Systems Synthesis Department of Automation Key Laboratory of System Control and Information Processing of Ministry of Education Shanghai200240 China Tongji University School of Computer Science and Technology National Pilot Software Engineering School with Chinese Characteristics Shanghai201804 China

Traditional LiDAR SLAM approaches prioritize localization over mapping, yet high-precision dense maps are essential for numerous applications involving intelligent agents. Recent advancements have introduced methods leveraging neural fields to enhance mapping capabilities;however, these approaches still face several limitations. Firstly, concerning scene representation, they typically employ neural fields with high-dimensional features and multi-layer perceptron decoders utilizing non-continuous activation functions. This results in low learning efficiency and challenges in capturing high-frequency signals. Secondly, in terms of scene organization, these methods often treat the entire scene as a singular neural field, leading to inefficiencies, inflexibility, and difficulties in rectifying accumulated errors when mapping large-scale environments over extended periods. To tackle the first issue, we propose a lightweight continuous SDF regression approach by encoding the scene in single-valued embeddings and decoding SDF values from a Kolmogorov-Arnold Network. By minimizing discrepancies in measuring range, sampling distance, and decoded SDF values, we facilitate iterative frame-to-model tracking and bundle adjustment neural mapping. To mitigate the second challenge, we propose structuring the whole scene into multiple neural SDF submaps. By establishing node-node, node-submap, and loop closure constraints into a global pose graph, the system can create dense neural maps with global consistency across large-scale scenes. Experimental evaluations in both real-world and simulated settings indicate that our system achieves superior mapping completeness and accuracy, enhanced learning efficiency, reduced memory consumption, and greater flexibility compared to its counterparts. © 2025 IEEE.

关键词： Mapping

Joint Optimization of Wireless Charging Station Location and Operation Scheduling for Electric Buses under Uncertain Demand

学校读者我要写书评

暂无评论

IEEE Access 2025年 13卷 91765-91776页

作者： Li, Jiacheng Noto, Masato Zhang, Yang Guo, Jia Kanagawa University Department of Applied Systems and Mathematics Yokohama Japan Kanagawa University Department of Computer Science Yokohama Japan Hubei Key Laboratory of Digital Finance Innovation Hubei University of Economics Wuhan China Hosei University Graduate School of Computer and Information Sciences Tokyo Japan Hubei University of Economics Japan School of Information Engineering Hubei Wuhan430205 China

In the context of electric bus route operations, the introduction of wireless charging facilities significantly impacts the operational scheduling of bus fleets. This study focuses on a single bus route and proposes a joint optimization model for determining departure frequencies and charging station locations under uncertain demand conditions. The model aims to minimize both user travel costs and transit operation costs by jointly optimizing departure schedules across different time periods, electric bus assignments, and wireless charging station placement. An adaptive genetic algorithm (AGA) is developed to solve the model efficiently. A case study based on a real-world bus route in China demonstrates that the proposed strategy not only improves system performance but also yields tangible benefits—reducing the required number of electric buses by 29.4%, shortening departure intervals by 12.5%, and lowering daily operational costs by 10.5%. © 2013 IEEE.

关键词： Genetic algorithms

A weakly supervised pavement crack segmentation based on adversarial learning and transformers

学校读者我要写书评

暂无评论

Multimedia systems 2025年第3期31卷 1-21页

作者： Apedo, Yvon Tao, Huanjie School of Computer Science Northwestern Polytechnical University Xi’an People’s Republic of China Engineering and Research Center of Embedded Systems Integration (Northwestern Polytechnical University) Ministry of Education Xi’an People’s Republic of China National Engineering Laboratory for Integrated Aero-Space-Ground-Ocean Big Data Application Technology Xi’an People’s Republic of China

While pixel-level crack segmentation has demonstrated significant potential in infrastructure inspections, its reliance on detailed annotations poses challenges for widespread adoption due to the time-consuming and expensive nature of such labeling. In response, weakly supervised crack segmentation has garnered attention as it eliminates the need for pixel-level annotations. However, existing methods, primarily based on class activation maps (CAM), involve complex training processes and suffer from poor performance due to misalignment between CAM-generated labels and the true target in the image. To address these challenges, we propose a weakly supervised approach for crack segmentation that simultaneously generates synthetic crack images and performs segmentation through adversarial learning. Unlike traditional methods that rely on coarse labels, our approach leverages synthetic crack images with corresponding labels, effectively eliminating issues of misalignment and noisy pseudo-labels. Our method introduces an encoder-decoder architecture for the segmentation model, incorporating a Transformer-based Feature Enhancement module (TFE) in the encoder. This module is specifically designed to capture long-range dependencies and efficiently integrate both high- and low-level features. Additionally, the model includes a Hilo block to extract both high- and low-frequency information from the images, along with a Progressive Shrinking Decoder (PSD) to aggregate and refine adjacent feature maps. Extensive experiments were conducted, and our model achieved an ODS of 67.12% on the CrackForest dataset, 60.11% on the Crack500 dataset, and 44.80% on the AEL dataset. These results demonstrate that our model outperforms several existing weakly supervised pavement segmentation methods across these datasets.

关键词：

MB-TaylorFormer V2: Improved Multi-branch Linear Transformer Expanded by Taylor Formula for Image Restoration

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Jin, Zhi Qiu, Yuwei Zhang, Kaihao Li, Hongdong Luo, Wenhan School of Intelligent Systems Engineering Sun Yat-sen University Shenzhen Campus Guangdong Shenzhen518107 China Guangdong Provincial Key Laboratory of Fire Science and Technology Guangzhou510006 China School of Intelligent Systems Engineering Sun Yat-sen University Shenzhen Campus Guangdong Shenzhen518107 China School of Computer Science and Technology Harbin Institute of Technology Shenzhen Campus Guangdong Shenzhen518055 China School of Engineering and Computer Science Australian National University CanberraACT2600 Australia Hong Kong University of Science and Technology Clear Water Bay Hong Kong

Recently, Transformer networks have demonstrated outstanding performance in the field of image restoration due to the global receptive field and adaptability to input. However, the quadratic computational complexity of Softmax-attention poses a significant limitation on its extensive application in image restoration tasks, particularly for high-resolution images. To tackle this challenge, we propose a novel variant of the Transformer. This variant leverages the Taylor expansion to approximate the Softmax-attention and utilizes the concept of norm-preserving mapping to approximate the remainder of the first-order Taylor expansion, resulting in a linear computational complexity. Moreover, we introduce a multi-branch architecture featuring multi-scale patch embedding into the proposed Transformer, which has four distinct advantages: 1) various sizes of the receptive field;2) multi-level semantic information;3) flexible shapes of the receptive field;4) accelerated training and inference speed. Hence, the proposed model, named the second version of Taylor formula expansion-based Transformer (for short MB-TaylorFormer V2) has the capability to concurrently process coarse-to-fine features, capture long-distance pixel interactions with limited computational cost, and improve the approximation of the Taylor expansion remainder. Experimental results across diverse image restoration benchmarks demonstrate that MB-TaylorFormer V2 achieves state-of-the-art performance in multiple image restoration tasks, such as image dehazing, deraining, desnowing, motion deblurring, and denoising, with very little computational overhead. The source code is available at https://***/FVL2020/MB-TaylorFormerV2. Copyright © 2025, The Authors. All rights reserved.

关键词： NP-hard

HEART: Achieving Timely Multi-Model Training for Vehicle-Edge-Cloud-Integrated Hierarchical Federated Learning

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Yang, Xiaohong Liwang, Minghui Wang, Xianbin Cheng, Zhipeng Hosseinalipour, Seyyedali Dai, Huaiyu Jiao, Zhenzhen School of Informatics Xiamen University Fujian China Department of Control Science and Engineering The National Key Laboratory of Autonomous Intelligent Unmanned Systems Japan Frontiers Science Center for Intelligent Autonomous Systems Ministry of Education Tongji University Shanghai China Department of Electrical and Computer Engineering Western University ON Canada School of Future Science and Engineering Soochow University Suzhou China Department of Electrical Engineering University at Buffalo-SUNY BuffaloNY United States Department of Electrical and Computer Engineering NC State University RaleighNC United States iF-Labs Beijing Teleinfo Technology Company Ltd. CAICT Beijing China

The rapid growth of AI-enabled Internet of Vehicles (IoV) calls for efficient machine learning (ML) solutions that can handle high vehicular mobility and decentralized data. This has motivated the emergence of Hierarchical Federated Learning over vehicle-edge-cloud architectures (VEC-HFL). Nevertheless, one aspect which is underexplored in the literature on VEC-HFL is that vehicles often need to execute multiple ML tasks simultaneously, where this multi-model training environment introduces crucial challenges. First, improper aggregation rules can lead to model obsolescence and prolonged training times. Second, vehicular mobility may result in inefficient data utilization by preventing the vehicles from returning their models to the network edge. Third, achieving a balanced resource allocation across diverse tasks becomes of paramount importance as it majorly affects the effectiveness of collaborative training. We take one of the first steps towards addressing these challenges via proposing a framework for multi-model training in dynamic VEC-HFL with the goal of minimizing global training latency while ensuring balanced training across various tasks—a problem that turns out to be NP-hard. To facilitate timely model training, we introduce a hybrid synchronous-asynchronous aggregation rule. Building on this, we present a novel method called Hybrid Evolutionary And gReedy allocaTion (HEART). The framework operates in two stages: first, it achieves balanced task scheduling through a hybrid heuristic approach that combines improved Particle Swarm Optimization (PSO) and Genetic Algorithms (GA);second, it employs a low-complexity greedy algorithm to determine the training priority of assigned tasks on vehicles. Experiments on real-world datasets demonstrate the superiority of HEART over existing methods. © 2025, CC BY-NC-ND.

关键词： Federated learning

PACF: Prototype Augmented Compact Features for Improving Domain Adaptive Object Detection

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Liu, Chenguang Feng, Yongchao Zhang, Yanan Liu, Qingjie Wang, Yunhong State Key Laboratory of Virtual Reality Technology and Systems Beihang University Beijing100191 China School of Computer Science and Information Engineering Hefei University of Technology Hefei230601 China Hangzhou Innovation Institute Beihang University Hangzhou310051 China

In recent years, there has been significant advancement in object detection. However, applying off-the-shelf detectors to a new domain leads to significant performance drop, caused by the domain gap. These detectors exhibit higher-variance class-conditional distributions in the target domain than that in the source domain, along with mean shift. To address this problem, we propose the Prototype Augmented Compact Features (PACF) framework to regularize the distribution of intra-class features. Specifically, we provide an in-depth theoretical analysis on the lower bound of the target features-related likelihood and derive the prototype cross entropy loss to further calibrate the distribution of target RoI features. Furthermore, a mutual regularization strategy is designed to enable the linear and prototype-based classifiers to learn from each other, promoting feature compactness while enhancing discriminability. Thanks to this PACF framework, we have obtained a more compact cross-domain feature space, within which the variance of the target features’ class-conditional distributions has significantly decreased, and the class-mean shift between the two domains has also been further reduced. The results on different adaptation settings are state-of-the-art, which demonstrate the board applicability and effectiveness of the proposed approach. Copyright © 2025, The Authors. All rights reserved.

关键词： Object detection

Real-Time Student Behavior Analysis via YOLOv5 with Coordinate Attention 9th

学校读者我要写书评

暂无评论

Real-Time Student Behavior Analysis via YOLOv5 with Coordina...

9th International Symposium on Artificial Intelligence and Robotics, ISAIR 2024

作者： Ma, Shuai Chang, Heyou Yang, Xinru Chen, Quan Han, Jian Key Laboratory of Intelligent Information Processing Nanjing XiaoZhuang University Nanjing China School of Information and Engineering Nanjing XiaoZhuang University Nanjing China School of Computer Science Nanjing University of Information Science and Technology Nanjing China Jiangsu Key Laboratory of Image and Video Understanding for Social Safety Nanjing University of Science and Technology Nanjing China Key Laboratory of Intelligent Perception and Systems for High-Dimensional Information of Ministry of Education Nanjing University of Science and Technology Nanjing China

ISBN: (纸本)9789819629138

The detection and recognition of student behavior play a pivotal role in the context of smart classrooms. However, conventional methods often encounter performance degradation due to challenges such as occlusion, data diversity, and complexity. To effectively addresses these issues, we propose a method that integrates YOLOv5 with coordinate attention mechanism and GhostConv for analyzing students' behavior. Specifically, the coordinate attention module enhances the localization ability of the proposed method by addressing problems related to target scale variations and uneven distribution. Additionally, we employ GhostConv to replace conventional convolutional layers in YOLOv5, reducing parameters while improving computational efficiency. To verify the validity of our method, a self-built challenging dataset for real teaching scenarios and Citypersons dataset are selected for the experiment. In comparison to the original YOLOv5, our method has achieved a 14.8% and 5.7% increase in mean average precision rates, with a 19.0% improvement in frames per second. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Students

LLM4WM: Adapting LLM for Wireless Multi-Tasking

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Liu, Xuanyu Gao, Shijian Liu, Boxun Cheng, Xiang Yang, Liuqing State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Peking University Beijing100871 China Guangzhou511400 China Guangzhou511400 China Department of Electronic and Computer Engineering The Department of Civil and Environmental Engineering The Hong Kong University of Science and Technology Hong Kong

The wireless channel is fundamental to communication, encompassing numerous tasks collectively referred to as channel-associated tasks. These tasks can leverage joint learning based on channel characteristics to share representations and enhance system design. To capitalize on this advantage, LLM4WM is proposed—a large language model (LLM) multi-task fine-tuning framework specifically tailored for channel-associated tasks. This framework utilizes a Mixture of Experts with Low-Rank Adaptation (MoE-LoRA) approach for multi-task fine-tuning, enabling the transfer of the pre-trained LLM’s general knowledge to these tasks. Given the unique characteristics of wireless channel data, preprocessing modules, adapter modules, and multi-task output layers are designed to align the channel data with the LLM’s semantic feature space. Experiments on a channel-associated multi-task dataset demonstrate that LLM4WM outperforms existing methodologies in both full-sample and few-shot evaluations, owing to its robust multi-task joint modeling and transfer learning capabilities. Copyright © 2025, The Authors. All rights reserved.

关键词： Semantics