检索结果-内蒙古大学图书馆

An efficient and balanced feature pyramid network: shared-FPN

Optoelectronics Letters 2025年

作者： Tian Sulong Qin Longfei Pang Wenchao Gao Ying Zhao Dexin School of Computer Science and Engineering， Tianjin University of Technology Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology， Tianjin University of Technology

In the field of object detection， it is challenging to achieve a balance between neck complexity and accuracy. To address this issue， we propose an efficient and decoupled neck module called Shared-FPN(Shared Feature Pyramid Network). Our design aims to effectively address the trade-off between these two factors. Neck modules are crucial for effectively fusing multi-scale information in neural network models. However， traditional neck designs， with their single structure， have struggled to significantly improve detection performance. In recent years， the incorporation of neural architecture search techniques in neck design has led to an increase in model complexity. This trade-off between model complexity and detection performance highlights the need for more efficient neck design approaches. Not only does Shared-FPN not increase the number of model parameters and FLOPs， but it also can be easily ported to any of the detection models. It improves on PAFPN by using transposed convolution with a large convolution kernel as the upsampling module， designing spatial pyramidal pooling-fast downsampling(SPPFD) based on shared pooling， and designing shared convolution as the right part module. To evaluate the performance of Shared-FPN in object detection tasks， we conducted experiments on object detection datasets. The results show that Shared-FPN achieved excellent performance across all sizes. In particular， on the VOC 2012 dataset， which contains mostly large objects， Shared-FPN's mAP was improved by 11.2% compared to FPN with YOLOX-s as the detector. On the COCO dataset， which contains mostly small objects， Shared-FPN's mAP was improved by 7.8% compared to FPN and 7.1% compared to PAFPN with Faster RCNN as the detector. Shared-FPN can be easily inserted into any of the detectors for better performance in various scenarios， such as small， medium or large objects.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Next-POI Recommendation via Spatial-Temporal Knowledge Graph Contrastive Learning and Trajectory Prompt

引用

IEEE Transactions on Knowledge and Data Engineering 2025年第6期37卷 3570-3582页

作者： Chen, Wei Huang, Haoyu Zhang, Zhiyu Wang, Tianyi Lin, Youfang Chang, Liang Wan, Huaiyu Guilin University of Electronic Technology Guangxi Key Lab of Trusted Software Guangxi Guilin541004 China Beijing Jiaotong University School of Computer Science and Technology Beijing100044 China Chinese University of Hong Kong Department of Computer Science and Engineering Ma Liu Shui Hong Kong Beijing Jiaotong University Beijing Key Laboratory of Traffic Data Mining and Embodied Intelligence Beijing100044 China

Next POI (Point-of-Interest) recommendation aims to forecast users' future movements based on their historical check-in trajectories, holding significant value in location-based services. Existing methods address trajectory data sparsity by integrating rich auxiliary information or using spatial-temporal knowledge graphs (STKGs), showing promising results. Yet, they face two main challenges: i) Due to the difficulty of transforming structured trajectory data into trajectory text describing users' spatial-temporal mobility, the powerful reasoning ability of pre-trained language models is rarely explored to enhance recommendation performance. ii) Methods based on STKG can introduce external knowledge inconsistent with user preferences, leading to the knowledge noise generated hampering the accuracy of recommendations. To this end, we propose a novel approach called STKG-PLM that integrates STKG contrastive learning and prompt pre-trained language model (PLM) to enhance the next POI recommendation. Specifically, we design a spatial-temporal trajectory prompt template that transforms structured trajectories into text corpus based on STKG, serving as the input of PLM to understand the movement pattern of users from coarse-grained and fine-grained perspectives. Additionally, we propose an STKG contrastive learning framework to mitigate the introduced knowledge noise. Extensive experiments on three real-world datasets demonstrate that STKG-PLM exhibits notable performance improvements over the state-of-the-art baseline methods. © 1989-2012 IEEE.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Profinder: Towards Professionals Recognition on Mobile Devices for Users with Cognitive Decline 18th

Profinder: Towards Professionals Recognition on Mobile Devi...

引用

18th International Conference on Wireless Artificial Intelligent Computing Systems and Applications, WASA 2024

作者： Wang, Yong Xiong, Yi Yang, Tianyu Shen, Yiran College of Computer Science and Technology Harbin Engineering University Harbin China Modeling and Emulation in E-Government National Engineering Laboratory Harbin China School of Software Shandong University Jinan China

ISBN: (纸本)9783031714696

As the global population continues to age, there is a concurrent rise in the number of individuals experiencing cognitive impairment and dementia, underscoring the critical necessity to address their hospice needs and quality of life concerns. Numerous studies have underscored the positive impact of exposure to outdoor environments on their overall condition and well-being. However, cognitive challenges often hinder their capacity to perform basic tasks independently, necessitating frequent assistance from onsite professionals. Complicating matters further, the complexity of outdoor settings often results in difficulties in identifying and recognizing personnel. This paper introduces Profinder, a system designed to assist individuals with cognitive impairment in identifying occupational personnel at specific outdoor locations and accessing essential assistance. Utilizing advanced deep learning techniques, Profinder seamlessly integrates foreground and background information, incorporates attention mechanisms, and harnesses siamese networks for efficient information decoupling, thereby achieving occupational recognition on smart mobile devices with an accuracy rate exceeding 90%. Additionally, a meticulously curated dataset dedicated to occupation recognition was devel0oped to facilitate and validate this research endeavor. By providing a robust auxiliary tool for individuals with cognitive impairment, this research endeavors to enhance their autonomy and overall quality of life, while also laying a solid foundation for further exploration and innovation in this burgeoning field, both within academia and industry. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Neurodegenerative diseases

来源：评论

学校读者我要写书评

暂无评论

Efficient and fine-grained access control with fully-hidden policies for cloud-enabled IoT

引用

Digital Communications and Networks 2025年第2期11卷 473-481页

作者： Li, Qi Liu, Gaozhan Zhang, Qianqian Han, Lidong Chen, Wei Li, Rui Xiong, Jinbo School of Computer Science Nanjing University of Posts and Telecommunications Nanjing210023 China Key Laboratory of Cryptography of Zhejiang Province Hangzhou Normal University Hangzhou311121 China Tongda College of Nanjing University of Posts and Telecommunications Yangzhou225127 China Henan Key Laboratory of Network Cryptography Technology Zhengzhou450001 China School of Computer Science and Technology Xidian University Xi'an710071 China State Key Laboratory of Tibetan Intelligent Information Processing and Application Qinghai Normal University Xining810008 China Faculty of Software Fujian Normal University Fuzhou350108 China

Ciphertext-Policy Attribute-Based Encryption (CP-ABE) enables fine-grained access control on ciphertexts, making it a promising approach for managing data stored in the cloud-enabled Internet of Things. But existing schemes often suffer from privacy breaches due to explicit attachment of access policies or partial hiding of critical attribute content. Additionally, resource-constrained IoT devices, especially those adopting wireless communication, frequently encounter affordability issues regarding decryption costs. In this paper, we propose an efficient and fine-grained access control scheme with fully hidden policies (named FHAC). FHAC conceals all attributes in the policy and utilizes bloom filters to efficiently locate them. A test phase before decryption is applied to assist authorized users in finding matches between their attributes and the access policy. Dictionary attacks are thwarted by providing unauthorized users with invalid values. The heavy computational overhead of both the test phase and most of the decryption phase is outsourced to two cloud servers. Additionally, users can verify the correctness of multiple outsourced decryption results simultaneously. Security analysis and performance comparisons demonstrate FHAC's effectiveness in protecting policy privacy and achieving efficient decryption. © 2024 Chongqing University of Posts and Telecommunications

关键词： Ciphertext

来源：评论

学校读者我要写书评

暂无评论

SSAST-Adapter: A Parameter-efficient Incremental Learning Algorithm for Underwater Acoustic Target Recognition

SSAST-Adapter: A Parameter-efficient Incremental Learning Al...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Zhu, Qian Xu, Qisheng Zhu, Boqing Gao, Zijian Zeng, Lingbin Xu, Kele College of Computer Science and Technology National University of Defense Technology Changsha China State Key Laboratory of Complex & Critical Software Environment Changsha China College of Meteorology and Oceanography National University of Defense Technology Changsha China

ISBN: (纸本)9798350368741

Underwater acoustic target recognition involves identifying and classifying targets in underwater environments using acoustic signals. In recent years, deep learning has made significant progress in this field. However, the models require the entire dataset to be available upfront, and classification categories must be predefined. In practical scenarios, new objects or species may appear in underwater environments over time, making it impractical to retrain a model from scratch each time new data is introduced. At the same time, training on new data inevitably leads to catastrophic forgetting of past data. To address these challenges, we propose a large-scale pre-training strategy combined with adapters to enable incremental learning without the need for complete retraining. To minimize the number of parameters during model fine-tuning, we employ a adapter structure at various network layers, reducing the number of trainable parameters to less than 2%. Experimental results demonstrate that our proposed method effectively reduces the need for full retraining by allowing the model to update in a resource-efficient manner using only new data, and the model has achieved high recognition accuracy in underwater acoustic target recognition tasks. © 2025 IEEE.

关键词： Adapter Audio Spectrum Transformer Incremental Learning Underwater Acoustics Target Recognition

来源：评论

学校读者我要写书评

暂无评论

A Hybrid Vectorized Merge Sort on ARM NEON 24th

A Hybrid Vectorized Merge Sort on ARM NEON

引用

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

作者： Zhou, Jincheng Zhang, Jin Zhang, Xiang Xiao, Tiaojie Ma, Di Gong, Chunye School of Computer and Communication Engineering Changsha University of Science and Technology Changsha410114 China College of Computer National University of Defense Technology Hunan Changsha410073 China Laboratory of Digitizing Software for Frontier Equipment National University of Defense Technology Changsha410073 China National Supercomputer Center in Tianjin Tianjin300457 China

ISBN: (纸本)9789819615506

Sorting algorithms are the most extensively researched topics in computer science and serve for numerous practical applications. Although various sorts have been proposed for efficiency, different architectures offer distinct flavors to the implementation of parallel sorting. In this paper, we propose a hybrid vectorized merge sort on ARM NEON, named NEON Merge Sort for short (NEON-MS). In detail, according to the granted register functions, we first identify the optimal register number to avoid the register-to-memory access, due to the write-back of intermediate outcomes. More importantly, following the generic merge sort framework that primarily uses sorting network for column sort and merging networks for three types of vectorized merge, we further improve their structures for high efficiency in an unified asymmetry way: 1) it makes the optimal sorting networks with few comparators become possible;2) hybrid implementation of both serial and vectorized merges incurs the pipeline with merge instructions highly interleaved. Experiments on a single FT2000+ core show that NEON-MS is 3.8 and 2.1 times faster than std::sort and boost::block_sort, respectively, on average. Additionally, as compared to the parallel version of the latter, NEON-MS gains an average speedup of 1.25. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Memory architecture

来源：评论

学校读者我要写书评

暂无评论

A difference vector angle dominance relation for expensive multi-objective optimization

引用

Swarm and Evolutionary Computation 2025年 95卷

作者： Yang, Cuicui Chen, Jing Ji, Junzhong Zhang, Xiaoyu Hao, Kangning Beijing Municipal Key Laboratory of Multimedia and Intelligent Software Technology College of Computer Science Beijing University of Technology Beijing China

For the latest two years, relation classification-based surrogate assisted evolutionary algorithms show good potential for solving expensive multi-objective optimization problems (EMOPs). However, the existing studies are still at the initial stage and lack specific research on the dominance relation. This paper proposes a difference vector angle dominance relation for EMOPs, which uses an angle threshold φ to control the selection pressure and is called DVAD-φ. The proposed DVAD-φ has adaptive selection pressure and considers the convergence and diversity of solutions when picking out superior solutions, which makes it beneficial to pick out promising solutions for expensive real FEs and reduce expensive real FEs. To be specific, we firstly give the definition of DVAD-φ that measures the superiority from one solution to another solution according to the angle threshold φ. Then, we deduce that there is monotonicity between the angle threshold φ and the number of non-dominated solutions in the sense of DVAD-φ. At last, we propose an adaptive determination strategy of angle threshold based on bisection to set proper pressure for picking out promising solutions for expensive real FEs. Experiments have been conducted on 23 test functions from two benchmark sets and one real-world problem. The experimental results have verified the effectiveness of DVAD-φ. © 2025

关键词： Optimization algorithms

来源：评论

学校读者我要写书评

暂无评论

Research on Microkernel Integrity Semantics Model and Formal Verification

引用

Chinese Journal of Electronics 2025年第1期23卷 43-48页

作者： Zhenjiang Qian Wei Liu Hao Huang State Key Laboratory for Novel Software Technology Nanjing University Nanjing China Department of Computer Science and Technology Nanjing University Nanjing China

Microkernel integrity is an important aspect of security for the whole microkernel system. Many of the research works on microkernel integrity focus on analysis and safeguards against the existing kernel attacks, and security enhancements for the vulnerabilities of system design and implementation aspects. The formal methods for operating system design and verification ensure the system's high level of security. The existing formalization work and research for operating system mainly focus on the code-level verification of program correctness. In this paper, we propose a formal abstraction model of microkernel in order to accurately describe the semantics of every kernel behavior, and achieve the description of the whole kernel. Based on the model we illustrate the microkernel integrity criterions and elaborate on the integrity mechanism. Meanwhile, we formally verify the completeness and consistency between the mechanism and the definition of the microkernel integrity. We use the self-implemented operating system VTOS (Verified trusted operating system) as an example to illustrate the design method for the microkernel integrity.

关键词： Design methodology Semantics Security Personnel Logic Kernel Protection System analysis and design Research and development Formal verification

来源：评论

学校读者我要写书评

暂无评论

MLPFormer: MLP-integrated transformer for colorectal histopathology whole slide image segmentation

引用

Neural Computing and Applications 2025年第6期37卷 4651-4661页

作者： Wang, Yuxuan Li, Dan Li, Xuechen Guo, Yan Zuo, Yanfei Shen, Linlin Department of Computer Science Wenzhou-Kean University Wenzhou China School of Electronics and Information Engineering Wuyi University Jiangmen China Department of Histo Pathology Shanghai Histo Pathology Diagnostic Center Shanghai China College of Computer Sicence and Software Engineering Shenzhen University Shenzhen China National Engineering Laboratory for Big Data System Computing Technology Shenzhen University Shenzhen China

Colorectal intraepithelial neoplasia is a precancerous lesion of colorectal cancer, which is mainly diagnosed using pathological images. According to the characteristics of lesions, precancerous lesions can be classified into four different grades, i.e., background, normal, low and high level. However, due to the continuity of lesion development, tissue in different stages has high similarity. Recently, visual transformers have achieved impressive results in different visual tasks, due to their capacity of capturing long-range dependencies. However, due to computational cost, transformer cannot well extract detailed feature, which is important for lesion grading. To accurately segment pathological tissues with high similarity, in this work, we embed the multi-head MLP mixer module in the transformer network to extract cell and edge information. Furthermore, we propose a hierarchical MLP decoder to facilitate multi-scale feature fusion. The proposed transformer, namely MLPFormer, achieves remarkable results against competitive baselines on the Histo-CRC Biopsy dataset. The experimental results demonstrate that MLPFormer significantly outperforms the competitive baselines, i.e., a 3% dice improvement is achieved over the SegFormer series. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Image segmentation

来源：评论

学校读者我要写书评

暂无评论

Learned Video Compression With Refined Adaptive Flow Pyramid And Coordinate-Aware Attention

Learned Video Compression With Refined Adaptive Flow Pyramid...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Huang, Qian Liu, Wenting Li, Xin Wang, Yiming College of Computer Science and Software Engineering Hohai University Nanjing China Key Laboratory of Water Big Data Technology Ministry of Water Resources Hohai University Nanjing China

ISBN: (纸本)9798350368741

In video compression, motion estimation and motion compensation are critical for achieving efficient encoding. Although the commonly used SpyNet and bilinear interpolation have contributed in improving the compression efficiency, they still have limitations. SpyNet often loses details and fails to fully utilize the feature extraction capabilities of deep networks. Furthermore, bilinear interpolation inherently attenuates high-frequency information, leading to frame blurring and distortion. In this paper, we propose a novel video compression algorithm. To overcome the limitations of SpyNet, we propose a refined adaptive flow pyramid network. This network uses a multi-scale feature pyramid to capture more details. Moreover, we use an iterative cost volume refinement engine that improves the feature representation of the network and iteratively improve the accuracy of motion estimation. In addition, to overcome the limitation of bilinear interpolation, we propose a coordinate-aware attention module, which captures more high-frequency information to improve the accuracy of motion compensation. Experimental results show that our method outperforms VTM-13.2 (LDP) in terms of PSNR. © 2025 IEEE.

关键词： coordinate network cross attention flow pyramid learned video compression

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：