检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Wang, Haoran Huai, Lian Li, Wenbin Qi, Lei Jiang, Xingqun Shi, Yinghuan National Key Laboratory for Novel Software Technology The National Institute of Healthcare Data Science Nanjing University Jiangsu Nanjing China National Key Laboratory for Novel Software Technology National Institute of Healthcare Data Science Nanjing University Nanjing Drum Tower Hospital Jiangsu Nanjing China National Key Laboratory for Novel Software Technology Nanjing University Jiangsu Nanjing China Lab School of Computer Science and Engineering Southeast University Jiangsu Nanjing China BOE Technology Group Co. Ltd. Beijing China

We have witnessed remarkable progress in foundation models in vision tasks. Currently, several recent works have utilized the segmenting anything model (SAM) to boost the segmentation performance in medical images, where most of them focus on training an adaptor for fine-tuning a large amount of pixel-wise annotated medical images following a fully supervised manner. In this paper, to reduce the labeling cost, we investigate a novel weakly-supervised SAM-based segmentation model, namely WeakMedSAM. Specifically, our proposed WeakMedSAM contains two modules: 1) to mitigate severe co-occurrence in medical images, a sub-class exploration module is introduced to learn accurate feature representations. 2) to improve the quality of the class activation maps, our prompt affinity mining module utilizes the prompt capability of SAM to obtain an affinity map for random-walk refinement. Our method can be applied to any SAM-like backbone, and we conduct experiments with SAMUS and EfficientSAM. The experimental results on three popularly-used benchmark datasets, i.e., BraTS 2019, AbdomenCT-1K, and MSD Cardiac dataset, show the promising results of our proposed WeakMedSAM. Our code is available at https://***/wanghr64/WeakMedSAM. Copyright © 2025, The Authors. All rights reserved.

关键词： Medical imaging

来源：评论

学校读者我要写书评

暂无评论

PIP: Perturbation-based Iterative Pruning for Large Language Models

arXiv

引用

arXiv 2025年

作者： Cao, Yi Xu, Wei-Jie Shen, Yucheng Shi, Weijie Chan, Chi-Min Xu, Jiajie School of Computer Science and Technology Soochow University China State Key Laboratory for Novel Software Technology Nanjing University China Department of Computer Science and Engineering The Hong Kong University of Science and Technology Hong Kong

The rapid increase in the parameter counts of Large Language Models (LLMs), reaching billions or even trillions, presents significant challenges for their practical deployment, particularly in resource-constrained environments. To ease this issue, we propose PIP (Perturbation-based Iterative Pruning), a novel double-view structured pruning method to optimize LLMs, which combines information from two different views: the unperturbed view and the perturbed view. With the calculation of gradient differences, PIP iteratively prunes those that struggle to distinguish between these two views. Our experiments show that PIP reduces the parameter count by approximately 20% while retaining over 85% of the original model’s accuracy across varied benchmarks. In some cases, the performance of the pruned model is within 5% of the unpruned version, demonstrating PIP’s ability to preserve key aspects of model effectiveness. Moreover, PIP consistently outperforms existing state-of-the-art (SOTA) structured pruning methods, establishing it as a leading technique for optimizing LLMs in environments with constrained resources. Our code is available at: https://***/caoyiiiiii/PIP. Copyright © 2025, The Authors. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Poster Abstract: IMU-assisted Image Stitching for Scenes with Obstructions Based on Camera Motion Sensing 25

Poster Abstract: IMU-assisted Image Stitching for Scenes wit...

引用

Proceedings of the 23rd ACM Conference on Embedded Networked Sensor Systems

作者： Saibing Han Yanling Bu Yanchao Zhao Lei Xie College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Nanjing China State Key Lab. for Novel Software Technology Nanjing University Nanjing China

ISBN: (纸本)9798400714795

Image stitching is often affected by obstructions such as pedestrians and vehicles. Traditional methods often ignore obstructions or retain them, leading to artifacts. They also cause edge distortion due to fixed-perspective stitching. To address above issues, this poster presents IIS, an IMU-assisted Image Stitching method. It uses IMU data for perspective pre-alignment, projecting images to an intermediate perspective to reduce edge distortion. It simultaneously removes obstructions and mitigating their impact. A weighted target scene stitching strategy is integrated to further enhance stitching quality. IIS effectively reduces artifacts and maintains high computational efficiency.

关键词： image stitching

来源：评论

学校读者我要写书评

暂无评论

E²GO: Free Your Hands for Smartphone Interaction

引用

International Journal of Human-computer Interaction 2025年

作者： Yan, Shaoming Ju, Yuanliang Quan, Rong Tu, Huawei Liang, Dong MIIT Key Laboratory of Pattern Analysis and Machine Intelligence Collaborative Innovation Center of Novel Software Technology and Industrialization College of Computer Science and Technology Nanjing University of Aeronautics and Astronautics Nanjing China Department of Computer Science and Information Technology La Trobe University Melbourne Australia

Current eye-gaze interaction technologies for smartphones are considered inflexible, inaccurate, and power-hungry. These methods typically rely on hand involvement and accomplish partial interactions. In this paper, we propose a novel eye-gaze smartphone interaction method named Event-driven Eye-Gaze Operation ((Formula presented.)), which can realize comprehensive interaction using only eyes and gazes to cover various interaction types. Before the interaction, an anti-jitter gaze estimation method was exploited to stabilize human eye fixation and predict accurate and stable human gaze positions on smartphone screens to further explore refined time-dependent eye-gaze interactions. We also integrated an event-triggering mechanism in (Formula presented.) to significantly decrease its power consumption to deploy on smartphones. We have implemented the prototype of (Formula presented.) on different brands of smartphones and conducted a comprehensive user study to validate its efficacy, demonstrating (Formula presented.) ‘s superior smartphone control capabilities across various scenarios. Demo videos. © 2025 Taylor & Francis Group, LLC.

关键词： Smartphones

来源：评论

学校读者我要写书评

暂无评论

MoMa: A Modular Deep Learning Framework for Material Property Prediction

arXiv

引用

arXiv 2025年

作者： Wang, Botian Ouyang, Yawen Li, Yaohui Wang, Yiqun Cui, Haorui Zhang, Jianbing Wang, Xiaonan Ma, Wei-Ying Zhou, Hao Tsinghua University China Department of Computer Science and Technology Tsinghua University China School of Artificial Intelligence National Key Laboratory for Novel Software Technology Nanjing University China Department of Chemical Engineering Tsinghua University China

Deep learning methods for material property prediction have been widely explored to advance materials discovery. However, the prevailing pre-train then fine-tune paradigm often fails to address the inherent diversity and disparity of material tasks. To overcome these challenges, we introduce MoMa, a Modular framework for Materials that first trains specialized modules across a wide range of tasks and then adaptively composes synergistic modules tailored to each downstream scenario. Evaluation across 17 datasets demonstrates the superiority of MoMa, with a substantial 14% average improvement over the strongest baseline. Few-shot and continual learning experiments further highlight MoMa’s potential for real-world applications. Pioneering a new paradigm of modular material learning, MoMa will be open-sourced to foster broader community collaboration. © 2025, CC BY-NC-SA.

关键词： Adversarial machine learning

来源：评论

学校读者我要写书评

暂无评论

Probabilistic-Search and Neighbor-Density Based Doppler-Shift Acquisition in Space Communications

引用

IEEE Transactions on Communications 2025年

作者： Zhang, Zhaowei Du, Shuai Liu, Lin Liu, Hui Zuo, Jiakuo Miao, Wang Zhao, Haitao Nanjing University of Posts and Telecommunications School of Internet of Things China Nanjing University State Key Laboratory for Novel Software Technology China Jinling Institute of Technology School of Software Engineering China University of Exeter Department of Computer Science United Kingdom

In space communications, the signal's long-distance transmission between a flying transmitter-receiver pair encounters two obstacles: a huge path-loss and a high-speed movement. The huge path-loss results in a low signal-to-noise ratio (SNR) and the high-speed movement brings a dynamic Doppler-shift, thus posing a great challenge for Doppler-shift acquisition. Under the low SNR, the Doppler-shift acquisition has to accumulate many symbols in a long-time period. However, during this period, the dynamic Doppler-shift disperses all these symbols' total energy over a wide range, thus causing the energy dispersion problem. To address this problem, we propose a probabilistic-search and neighbor-density (PSND) scheme, where the probabilistic-search considers the element's amplitude and sacrifices some redundant signal elements for a narrower search-range, while the neighbordensity sums the signal element's all neighbors for a larger signal energy. The PSND scheme includes two algorithms: the Generic PSND and Iterative PSND. The Generic PSND algorithm first selects some search-elements with their probabilities to construct a probabilistic-search-range, then selects some density-elements with their neighbor-elements to derive the neighbor-densities, and finally searches the largest neighbor-density to obtain the acquisition result. Built upon the Generic PSND algorithm above, the Iterative PSND algorithm iteratively updates the searchelements to further strengthen the neighbor-density in a narrower probabilistic-search-range. Moreover, the simulations results have demonstrated our PSND scheme's higher acquisition probability, as compared with the existing schemes. © 1972-2012 IEEE.

关键词： Signal to noise ratio

来源：评论

学校读者我要写书评

暂无评论

LLFA: Fusing Global Illumination and Local Priors for Low-Light Face Image Enhancement with Adaptor

LLFA: Fusing Global Illumination and Local Priors for Low-Li...

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Ziqian Shao Tao Wang Kaihao Zhang Danhuai Zhao Tong Lu State Key Lab for Novel Software Technology Nanjing University Nanjing China School of Computer Science and Technology Harbin Institute of Technology Shenzhen Shenzhen China China Mobile Zijin Innovation Insititute Nanjing China

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Low-light image enhancement problem has been widely studied. However, most existing methods do not perform well on low-light face images due to no specific facial characteristic considerations. We first create large-scale low-light face datasets with synthesized and real-world images to address the absence of suitable datasets. Our experiments show that existing LLIE and face restoration methods are limited in enhancing low-light face images. To overcome these challenges, we propose a novel framework, the Low-Light Face Adaptor (LLFA), featuring an auxiliary encoder and an adaptor module. The encoder captures global illumination information, while the adaptor module adaptively fuses this information with high-quality priors. We also introduce a joint learning strategy that optimizes the model by simultaneously learning face priors and the enhancement process. Comprehensive experiments demonstrate that LLFA significantly outperforms state-of-the-art methods.

关键词： Training Adaptation models Fuses Lighting Speech enhancement Signal processing Image restoration Faces Image enhancement Facial features

来源：评论

学校读者我要写书评

暂无评论

Question Selection for Multi-Modal Code Search Synthesis using Probabilistic Version Spaces

引用

IEEE Transactions on software Engineering 2025年

作者： Wu, Jiarong Jiang, Yanyan Wei, Lili Xu, Congying Cheung, Shing-Chi Xu, Chang The Hong Kong University of Science and Technology Department of Computer Science and Engineering Hong Kong McGill University Department of Electrical and Computer Engineering Montreal Canada Nanjing University State Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing China

Searching the occurrences of specific code patterns (code search) is a common task in software engineering, and programming by example (PBE) techniques have been applied to ease customizing code patterns. However, previous PBE tools only synthesize programs meeting the input-output examples, which may not always align with the user intent. To bridge this gap, this paper proposes Excalibur, a multi-modal (example and natural language description) and interactive synthesizer for code search. Excalibur ensures that the generated programs are correct for the provided examples (soundness) and include the user-intended program (bounded completeness). Furthermore, Excalibur helps the user identify the user-intended program through question-answer interaction. To minimize the required interaction efforts, question selection is crucial. To improve question selection for code search, we propose probabilistic version spaces (ProbVS), in which the user-intended program’s probability is high and others are low. ProbVS combines traditional version spaces for compactly representing extensive programs and large language models (on the user-provided natural language description) for adjusting programs’ probabilities to align with users’ intents. Extensive experiments on a benchmark of 44 tasks demonstrated the effectiveness of Excalibur and ProbVS and demystified how ProbVS affects probability distributions and how the configurable parameters affect ProbVS. © 1976-2012 IEEE.

关键词： Normal distribution

来源：评论

学校读者我要写书评

暂无评论

ExpertDRL: Request Dispatching and Instance Configuration for Serverless Edge Inference with Foundation Models

引用

IEEE Transactions on Mobile Computing 2025年

作者： Zeng, Yue Zhou, Junlong Ye, Baoliu Qu, Zhihao Guo, Song Gong, Tianjian Li, Pan Nanjing University of Science and Technology Department of Computer Science and Engineering Jiangsu Nanjing210094 China Nanjing University State Key Laboratory for Novel Software Technology Department of Computer Science and Technology Nanjing210023 China Hohai University Key Laboratory of Water Big Data Technology of Ministry of Water Resources Nanjing211100 China Hong Kong University of Science and Technology Hong Kong Southwest University College of Electronic and Information Engineering Chongqing400715 China

The prevalence of the pre-training & fine-tuning paradigm enables machine learning models to quickly adapt to various downstream tasks by fine-tuning pre-trained foundation models (FMs), greatly facilitating various IoT applications that rely on model inference in dynamic edge serverless environments. Efficiently dispatching inference requests and configuring instances to batch inference requests can significantly enhance resource efficiency. However, existing serverless inference solutions are tailored for traditional models, make coarse-grained request dispatching and instance configuration decisions, fail to exploit the shared model backbone characteristics of the FM and capture delayed rewards in dynamic environments, and ignore communication latency between edge sites, resulting in high costs and constraint violations. In this paper, we leverage our insight that fine-grained batch inference requests can effectively exploit the shared model backbone feature of FM to save monetary costs. We propose an algorithm that incorporates deep reinforcement learning (DRL) and expert intervention for fine-grained request dispatching and instance configuration, where the DRL component outputs fractional solutions as guidance, while the expert intervention module integrates our insights—batching reduces monetary costs at the expense of increased inference latency, whereas higher configurations shorten inference latency. This module rounds fractional solutions and adjusts instance configurations to search for optimal solutions while satisfying constraints, with theoretical guarantees rigorously proved. Finally, we conducted our experiments on an OpenFaas-based platform and simulator, and extensive trace-driven evaluation results show that ExpertDRL can save costs by up to 85.14% and improve request acceptance ratio by up to 26.93%, compared to the state-of-the-art solution. © 2002-2012 IEEE.

关键词： Deep reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

3D Gaussian Splatting: Survey, Technologies, Challenges, and Opportunities

引用

IEEE Transactions on Circuits and Systems for Video technology 2025年

作者： Bao, Yanqi Ding, Tianyu Huo, Jing Liu, Yaoli Li, Yuxin Li, Wenbin Gao, Yang Luo, Jiebo Nanjing University State Key Laboratory for Novel Software Technology 210023 China Microsoft Corporation Applied Sciences Group Redmond United States University of Rochester Department of Computer Science United States

3D Gaussian Splatting (3DGS) has emerged as a prominent technique with the potential to become a mainstream method for 3D representations. It can effectively transform multi-view images into explicit 3D Gaussian through efficient training, and achieve real-time rendering of novel views. This survey aims to analyze existing 3DGS-related works from multiple intersecting perspectives, including related tasks, technologies, challenges, and opportunities. The primary objective is to provide newcomers with a rapid understanding of the field and to assist researchers in methodically organizing existing technologies and challenges. Specifically, we delve into the optimization, application, and extension of 3DGS, categorizing them based on their focuses or motivations. Additionally, we summarize and classify nine types of technical modules and corresponding improvements identified in existing works. Based on these analyses, we further examine the common challenges and technologies across various tasks, proposing potential research opportunities. © 1991-2012 IEEE.

关键词： 3D modeling

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：