检索结果-内蒙古大学图书馆

CRRNet: Channel Relation Reasoning Network for Salient Object Detection 3rd

学校读者我要写书评

暂无评论

CRRNet: Channel Relation Reasoning Network for Salient Obje...

3rd China intelligent Robotics Annual Conference, CCF CIRAC 2022

作者： Gao, Shuyong Xing, Haozhe Zhang, Chenglong Zhang, Wenqiang Shanghai Key Laboratory of Intelligent Information Processing Fudan University Shanghai200433 China School of Computer Science Fudan University Shanghai200433 China Academy for Engineering and Technology Fudan University Shanghai200433 China Yiwu Research Institute of Fudan University Chengbei Road Zhejiang Yiwu City322000 China

ISBN: (纸本)9789819903009

Channel map matters for salient object detection. Effectively exploring the relationship between channels can enhance the relevant channel maps and infer a better saliency map. The existing channel map enhancement methods focus on learning the weights to re-weight the channel maps or using the similarity between the channels to combine the channel maps and enhance them. We propose a Channel Relation Reasoning Module (CRRM) based on graph convolution, inferring the semantic relations between channels and enhancing the channel maps according to the mutual relations. First, we carefully designed a multi-level complementary feature extraction module, which is used to extract multi-level edge and salient object internal features simultaneously. Then, the rich multi-level edge and salient features are fed to CRRM which projects the features into the hidden space and uses the graph convolution to establish the semantic relationship between the channels to enhance the feature expression. Finally, two stacked convolution layers can produce high-quality saliency maps. Extensive experiments on multiple datasets show the effectiveness of our model. © 2023, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.

关键词： Convolution

Inference-Time Decontamination: Reusing Leaked Benchmarks for Large Language Model Evaluation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhu, Qin Cheng, Qinyuan Peng, Runyu Li, Xiaonan Liu, Tengxiao Peng, Ru Qiu, Xipeng Huang, Xuanjing School of Computer Science Fudan University China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China College of Computer Science and Technology Zhejiang University China

The training process of large language models (LLMs) often involves varying degrees of test data contamination (Yang et al., 2023b). Although current LLMs are achieving increasingly better performance on various benchmarks, their performance in practical applications does not always match their benchmark results. Leakage of benchmarks can prevent the accurate assessment of LLMs’ true performance. However, constructing new benchmarks is costly, labor-intensive and still carries the risk of leakage. Therefore, in this paper, we ask the question "Can we reuse these leaked benchmarks for LLM evaluation?" We propose Inference-Time Decontamination (ITD) to address this issue by detecting and rewriting leaked samples without altering their difficulties. ITD can mitigate performance inflation caused by memorizing leaked benchmarks. Our proof-of-concept experiments demonstrate that ITD reduces inflated accuracy by 22.9% on GSM8K and 19.0% on MMLU. On MMLU, using Inference-time Decontamination can lead to a decrease in the results of Phi3 and Mistral by 6.7% and 3.6% respectively. We hope that ITD can provide more truthful evaluation results for large language models. Copyright © 2024, The Authors. All rights reserved.

关键词： Decontamination

Forget less,count better:a domain-incremental self-distillation learning benchmark for lifelong crowd counting

学校读者我要写书评

暂无评论

Frontiers of information technology & Electronic Engineering 2023年第2期24卷 187-202页

作者： Jiaqi GAO Jingqi LI Hongming SHAN Yanyun QU James ZWANG Fei-Yue WANG Junping ZHANG Shanghai Key Laboratory of Intelligent Information Processing School of Computer ScienceFudan UniversityShanghai 200433China Institute of Science and Technology for Brain-inspired Intelligence Fudan UniversityShanghai 200433China Shanghai Center for Brain Science and Brain-inspired Technology Shanghai 201210China School of Information Science and Technology Xiamen UniversityXiamen 361005China College of Information Sciences and Technology the Pennsylvania State UniversityUniversity ParkPA 16802USA State Key Laboratory of Management and Control for Complex Systems Institute of AutomationChinese Academy of SciencesBeijing 100190China

Crowd counting has important applications in public safety and pandemic control.A robust and practical crowd counting system has to be capable of continuously learning with the newly incoming domain data in real-world scenarios instead of fitting one domain ***-the-shelf methods have some drawbacks when handling multiple domains:(1)the models will achieve limited performance(even drop dramatically)among old domains after training images from new domains due to the discrepancies in intrinsic data distributions from various domains,which is called catastrophic forgetting;(2)the well-trained model in a specific domain achieves imperfect performance among other unseen domains because of domain shift;(3)it leads to linearly increasing storage overhead,either mixing all the data for training or simply training dozens of separate models for different domains when new ones are *** overcome these issues,we investigate a new crowd counting task in incremental domain training setting called lifelong crowd *** goal is to alleviate catastrophic forgetting and improve the generalization ability using a single model updated by the incremental ***,we propose a self-distillation learning framework as a benchmark(forget less,count better,or FLCB)for lifelong crowd counting,which helps the model leverage previous meaningful knowledge in a sustainable manner for better crowd counting to mitigate the forgetting when new data arrive.A new quantitative metric,normalized Backward Transfer(nBwT),is developed to evaluate the forgetting degree of the model in the lifelong learning *** experimental results demonstrate the superiority of our proposed benchmark in achieving a low catastrophic forgetting degree and strong generalization ability.

关键词： Crowd counting Knowledge distillation Lifelong learning

RBDQ: A Reliable LLM-based Text-to-SQL System for Business Data Queries 25

学校读者我要写书评

暂无评论

RBDQ: A Reliable LLM-based Text-to-SQL System for Business D...

Companion Proceedings of the ACM on Web Conference 2025

作者： Fenglin Bi Dongdong Cao Zhiyu Wang Yang Chen Fangliang Zhao Tao Hu Zhi Li Yanbin Zhang Wei Wang School of Data Science and Engineering East China Normal University Shanghai China ByteDance Inc Shanghai China School of Computer Science Fudan University Shanghai China and Shanghai Key Lab of Intelligent Information Processing Shanghai China

ISBN: (纸本)9798400713316

Using large language models (LLMs) to convert natural language (NL) into SQL simplifies data access for users by allowing them to use everyday language. However, business departments often distrust LLM-based text-to-SQL systems due to the probabilistic nature of SQL generation, which can result in incorrect but executable SQL queries caused by model hallucinations. This leads to significant concerns regarding the accuracy and reliability of the queried data. In this paper, we present RBDQ, a novel LLM-based text-to-SQL system designed to address the unique challenges of business data queries. RBDQ innovatively introduces the Hierarchical Metrics Query Method and integrates advanced Retrieval-Augmented Generation (RAG) methods along with a self-reflection mechanism to tackle these challenges. RBDQ effectively meets the requirements of business metric queries in real-world scenarios. Currently implemented in the Quality Assurance department at ByteDance, RBDQ has significantly improved operational efficiency and query flexibility. Our experiments demonstrate the system's effectiveness, achieving an Execution Accuracy of 96.20%.

关键词： business data queries

Global Convergence Guarantees of (A)GIST for a Family of Nonconvex Sparse Learning Problems

学校读者我要写书评

暂无评论

IEEE Transactions on Cybernetics 2022年第5期52卷 3276-3288页

作者： Zhang, Hengmin Qian, Feng Shang, Fanhua Du, Wenli Qian, Jianjun Yang, Jian East China University of Science and Technology Key Laboratory of Advanced Control and Optimization for Chemical Processes of Ministry of Education School of Information Science and Engineering Shanghai200237 China Shanghai Institute of Intelligent Science and Technology Tongji University Shanghai200092 China Xidian University Key Laboratory of Intelligent Perception and Image Understanding of Ministry of Education School of Artificial Intelligence Xi'an710071 China Peng Cheng Laboratory Shenzhen518066 China Nanjing University of Science and Technology Pca Laboratory Key Lab. of Intelligent Percept. and Syst. for High-Dimensional Information of Ministry of Education Nanjing210094 China Nanjing University of Science and Technology Jiangsu Key Laboratory of Image and Video Understanding for Social Security School of Computer Science and Engineering Nanjing210094 China

In recent years, most of the studies have shown that the generalized iterated shrinkage thresholdings (GISTs) have become the commonly used first-order optimization algorithms in sparse learning problems. The nonconvex relaxations of the \ell _{0} -norm usually achieve better performance than the convex case (e.g., \ell _{1} -norm) since the former can achieve a nearly unbiased solver. To increase the calculation efficiency, this work further provides an accelerated GIST version, that is, AGIST, through the extrapolation-based acceleration technique, which can contribute to reduce the number of iterations when solving a family of nonconvex sparse learning problems. Besides, we present the algorithmic analysis, including both local and global convergence guarantees, as well as other intermediate results for the GIST and AGIST, denoted as (A)GIST, by virtue of the Kurdyka-Lojasiewica (KL) property and some milder assumptions. Numerical experiments on both synthetic data and real-world databases can demonstrate that the convergence results of objective function accord to the theoretical properties and nonconvex sparse learning methods can achieve superior performance over some convex ones. © 2013 IEEE.

关键词： Optimization

Friedkin-Johnsen Model for Opinion Dynamics on Signed Graphs

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Zhou, Xiaotian Sun, Haoxin Xu, Wanyue Li, Wei Zhang, Zhongzhi Shanghai Key Laboratory of Intelligent Information Processing School of Computer Science Fudan University Shanghai200433 China The Academy for Engineering and Technology Fudan University Shanghai200433 China

A signed graph offers richer information than an unsigned graph, since it describes both collaborative and competitive relationships in social networks. In this paper, we study opinion dynamics on a signed graph, based on the Friedkin-Johnsen model. We first interpret the equilibrium opinion in terms of a defined random walk on an augmented signed graph, by representing the equilibrium opinion of every node as a combination of all nodes’ internal opinions, with the coefficient of the internal opinion for each node being the difference of two absorbing probabilities. We then quantify some relevant social phenomena and express them in terms of the 2 norms of vectors. We also design a nearly-linear time signed Laplacian solver for assessing these quantities, by establishing a connection between the absorbing probability of random walks on a signed graph and that on an associated unsigned graph. We further study the opinion optimization problem by changing the initial opinions of a fixed number of nodes, which can be optimally solved in cubic time. We provide a nearly-linear time algorithm with error guarantee to approximately solve the problem. Finally, we execute extensive experiments on sixteen real-life signed networks, which show that both of our algorithms are effective and efficient, and are scalable to massive graphs with over 20 million nodes. Copyright © 2024, The Authors. All rights reserved.

关键词： Clustering algorithms

Delving into Decision-based Black-box Attacks on Semantic Segmentation

学校读者我要写书评

暂无评论

arXiv 2024年

作者： Chen, Zhaoyu Shan, Zhengyang Chang, Jingwen Jiang, Kaixun Yang, Dingkang Cheng, Yiting Zhang, Wenqiang Shanghai Engineering Research Center of AI & Robotics Academy for Engineering & Technology Fudan University China Shanghai Key Lab of Intelligent Information Processing School of Computer Science Fudan University China

Semantic segmentation is a fundamental visual task that finds extensive deployment in applications with security-sensitive considerations. Nonetheless, recent work illustrates the adversarial vulnerability of semantic segmentation models to white-box attacks. However, its adversarial robustness against black-box attacks has not been fully explored. In this paper, we present the first exploration of black-box decision-based attacks on semantic segmentation. First, we analyze the challenges that semantic segmentation brings to decision-based attacks through the case study. Then, to address these challenges, we first propose a decision-based attack on semantic segmentation, called Discrete Linear Attack (DLA). Based on random search and proxy index, we utilize the discrete linear noises for perturbation exploration and calibration to achieve efficient attack efficiency. We conduct adversarial robustness evaluation on 5 models from Cityscapes and ADE20K under 8 attacks. DLA shows its formidable power on Cityscapes by dramatically reducing PSPNet’s mIoU from an impressive 77.83% to a mere 2.14% with just 50 queries. Copyright © 2024, The Authors. All rights reserved.

关键词： Semantics

Anomize: Better Open Vocabulary Video Anomaly Detection

学校读者我要写书评

暂无评论

arXiv 2025年

作者： Li, Fei Liu, Wenxuan Chen, Jingjing Zhang, Ruixu Wang, Yuran Zhong, Xian Wang, Zheng National Engineering Research Center for Multimedia Software School of Computer Science Wuhan University China Shanghai Key Lab of Intelligent Information Processing School of Computer Science Fudan University China State Key Laboratory for Multimedia Information Processing School of Computer Science Peking University China Hubei Key Laboratory of Transportation Internet of Things Wuhan University of Technology China

Open Vocabulary Video Anomaly Detection (OVVAD) seeks to detect and classify both base and novel anomalies. However, existing methods face two specific challenges related to novel anomalies. The first challenge is detection ambiguity, where the model struggles to assign accurate anomaly scores to unfamiliar anomalies. The second challenge is categorization confusion, where novel anomalies are often misclassified as visually similar base instances. To address these challenges, we explore supplementary information from multiple sources to mitigate detection ambiguity by leveraging multiple levels of visual data alongside matching textual information. Furthermore, we propose incorporating label relations to guide the encoding of new labels, thereby improving alignment between novel videos and their corresponding labels, which helps reduce categorization confusion. The resulting Anomize framework effectively tackles these issues, achieving superior performance on UCF-CRIME and XD-VIOLENCE datasets, demonstrating its effectiveness in OVVAD. Copyright © 2025, The Authors. All rights reserved.

关键词：

Improving Drum Source Separation with Temporal-Frequency Statistical Descriptors

学校读者我要写书评

暂无评论

Improving Drum Source Separation with Temporal-Frequency Sta...

IEEE International Conference on Multimedia and Expo (ICME)

作者： Si Li Jiaxing Liu Peilin Li Dichucheng Li Xinlu Liu Yongwei Gao Wei Li School of Computer Science and Technology Fudan University China School of Statistics and Information Shanghai University of International Business and Economics China Shanghai Key Laboratory of Intelligent Information Processing Fudan University China

ISBN: (数字)9798350390155

ISBN: (纸本)9798350390162

Drum Source Separation (DSS) aims to separate drum mixtures into individual drum sounds, such as kick and snare. Deep neural network methods have been successfully applied for source separation. However, due to the limited size of existing datasets and the strongly overlap of drums in frequency and time, these methods still have certain shortcomings. To address these challenges, we construct a large drum sound dataset and propose a novel training objective to improve performance of DSS task. The training objective leverages three temporal-frequency statistical descriptors (spectral centroid, spectral spread, and spectral flux) to separate drum sources. Our experimental results demonstrate that our method can make a SDR improvement of 0.98 dB on UNet and 1.07 dB on MERT. Furthermore, our method achieves consistent improvements in low-resource and cross-dataset scenarios. Our code and dataset are available at https://***/150042/Drum-Separation-TF.

关键词： Training Deep learning Adaptation models Time-frequency analysis Visualization Tracking loops Source separation