检索结果-内蒙古大学图书馆

Cross-domain object detection through coarse-to-fine feature adaptation

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Zheng, Yangtao Huang, Di Liu, Songtao Wang, Yunhong Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University State Key Laboratory of Software Development Environment Beihang University School of Computer Science and Engineering Beihang University Beijing100191 China

Recent years have witnessed great progress in deep learning based object detection. However, due to the domain shift problem, applying off-the-shelf detectors to an unseen domain leads to significant performance drop. To address such an issue, this paper proposes a novel coarseto-fine feature adaptation approach to cross-domain object detection. At the coarse-grained stage, different from the rough image-level or instance-level feature alignment used in the literature, foreground regions are extracted by adopting the attention mechanism, and aligned according to their marginal distributions via multi-layer adversarial learning in the common feature space. At the fine-grained stage, we conduct conditional distribution alignment of foregrounds by minimizing the distance of global prototypes with the same category but from different domains. Thanks to this coarse-to-fine feature adaptation, domain knowledge in foreground regions can be effectively transferred. Extensive experiments are carried out in various cross-domain detection scenarios. The results are state-of-the-art, which demonstrate the broad applicability and effectiveness of the proposed approach. Copyright © 2020, The Authors. All rights reserved.

关键词： Object recognition

Dynamic graph-based anomaly detection in the electrical grid

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Li, Shimiao Pandey, Amritanshu Hooi, Bryan Faloutsos, Christos Pileggi, Larry Department of Electrical and Computer Engineering Carnegie Mellon University PittsburghPA15213 United States School of Computing The Institute of Data Science in National University of Singapore Singapore Department of Computer Science Carnegie Mellon University PittsburghPA15213 United States

Given sensor readings over time from a power grid, how can we accurately detect when an anomaly occurs? A key part of achieving this goal is to use the network of power grid sensors to quickly detect, in real-time, when any unusual events, whether natural faults or malicious, occur on the power grid. Existing bad-data detectors in the industry lack the sophistication to robustly detect broad types of anomalies, especially those due to emerging cyber-attacks, since they operate on a single measurement snapshot of the grid at a time. New ML methods are more widely applicable, but generally do not consider the impact of topology change on sensor measurements and thus cannot accommodate regular topology adjustments in historical data. Hence, we propose DYNWATCH, a domain knowledge based and topology-aware algorithm for anomaly detection using sensors placed on a dynamic grid. Our approach is accurate, outperforming existing approaches by 20% or more (F-measure) in experiments;and fast, averaging less than 1.7ms per time tick per sensor on a 60K+ branch case using a laptop computer, and scaling linearly with the size of the graph. Copyright © 2020, The Authors. All rights reserved.

关键词： Anomaly detection

A Study of Current Trends Writer Identification in Large-scale Across Three World Major Languages with Retrieval Approaches

学校读者我要写书评

暂无评论

A Study of Current Trends Writer Identification in Large-sca...

International Conference on Interactive Digital Media (ICIDM)

作者： Gloria Jennis Tan Rosely Kumoi Mohd Shafry Mohd Rahim Tan Chi Wee Ghazali Sulong Universiti Malaysia Sarawak Sarawak Malaysia School of Computing Universiti Teknologi Malaysia Johor Malaysia Tunku Abdul Rahman University College Malaysia Centre for Cyber Security & Big Data Management & Science University Malaysia

ISBN: (数字)9781728149288

ISBN: (纸本)9781728142050

Recent developments in writer identification have led new directions, which is writer identification integrated with retrieval mechanism for boosting the performance applications of writer identification when a database is getting larger in size and retrieval of documents from expanding databases is becoming difficult. Issues on datasets used by previous studies directly impacts accuracy of the writer identification deteriorates as database size increases. Until now, in the research on writer identification researchers have been mainly interested in “writer identification without retrieval mechanism” approach. This paper provides a review on the state of the art for writer identification with retrieval mechanism throughout three major languages, namely English, Chinese, and Arabic. Up to date, truly little attention has been paid to the role of retrieval mechanism or documents in English language identification; the growth on Arabic and Chinese is far from satisfactory in comparison to its extensive usage. A search of the literature revealed few studies for writer identification with single retrieval mechanism. Seamless integration of such strategies in current forensic handwriting experience is yet unclear. A consistently reliable and efficient, fully automated writer recognition (AWR) and a powerful handwritten document retrieval system as an emerging technology are real challenging issues might be taken by upcoming researchers. It is still considered a fresh field and still has large room for research.

关键词： Handwriting recognition databases Forensics Media Market research Reliability Standards

The EMory BrEast imaging dataset (EMBED): A Racially Diverse, Granular dataset of 3.5M Screening and Diagnostic Mammograms

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Jeong, Jiwoong J. Vey, Brianna L. Reddy, Ananth Kim, Thomas Santos, Thiago Correa, Ramon Dutt, Raman Mosunjac, Marina Oprea-Ilies, Gabriela Smith, Geoffrey Woo, Minjae McAdams, Christopher R. Newell, Mary S. Banerjee, Imon Gichoya, Judy Trivedi, Hari Department of Computer Science Emory University 201 Dowman Drive AtlantaGA30322 United States Department of Radiology Emory University 1364 Clifton Rd AtlantaGA30322 United States College of Computing Georgia Institute of Technology 801 Atlantic Dr NW AtlantaGA30332 United States Department of Pathology Emory University 1364 Clifton Rd AtlantaGA30322 United States School of Data Science and Analytics Kennesaw State University 1000 Chastain Road KennesawGA30144 United States Department of Biomedical Informatics Emory University 100 Woodruff Circle AtlantaGA30322 United States

Developing and validating artificial intelligence models in medical imaging requires datasets that are large, granular, and diverse. To date, the majority of publicly available breast imaging datasets lack in one or more of these areas. Models trained on these data may therefore underperform on patient populations or pathologies that have not previously been encountered. The EMory BrEast imaging dataset (EMBED) addresses these gaps by providing 3650,000 2D and DBT screening and diagnostic mammograms for 116,000 women divided equally between White and African American patients. The dataset also contains 40,000 annotated lesions linked to structured imaging descriptors and 61 ground truth pathologic outcomes grouped into six severity classes. Our goal is to share this dataset with research partners to aid in development and validation of breast AI models that will serve all patients fairly and help decrease bias in medical AI. © 2022, CC BY.

关键词： Mammography

Pixel-wise dense detector for image inpainting

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Zhang, Ruisong Quan, Weize Wu, Baoyuan Li, Zhifeng Yan, Dong-Ming National Laboratory of Pattern Recognition Institute of Automation Chinese Academy of Sciences Beijing100190 China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China School of Data Science Chinese University of Hong Kong Shenzhen Hong Kong Secure Computing Lab of Big Data Shenzhen Research Institute of Big Data China Tencent AI Lab Shenzhen China

Recent GAN-based image inpainting approaches adopt an average strategy to discriminate the generated image and output a scalar, which inevitably lose the position information of visual artifacts. Moreover, the adversarial loss and reconstruction loss (e.g., `1 loss) are combined with tradeoff weights, which are also difficult to tune. In this paper, we propose a novel detection-based generative framework for image inpainting, which adopts the min-max strategy in an adversarial process. The generator follows an encoder-decoder architecture to fill the missing regions, and the detector using weakly supervised learning localizes the position of artifacts in a pixel-wise manner. Such position information makes the generator pay attention to artifacts and further enhance them. More importantly, we explicitly insert the output of the detector into the reconstruction loss with a weighting criterion, which balances the weight of the adversarial loss and reconstruction loss automatically rather than manual operation. Experiments on multiple public datasets show the superior performance of the proposed framework. The source code is available at https://***/Evergrow/GDN_Inpainting. Copyright © 2020, The Authors. All rights reserved.

关键词： Pixels

229TiP A phase II trial of a neural network-based treatment decision support tool in patients (pts) with refractory solid organ malignancies

学校读者我要写书评

暂无评论

Annals of Oncology 2024年 35卷 S307-S308页

作者： R.J. Walsh A. Jayagopal T.Z. Tan J. Pitt R. Sundar S.C. Lee B-C. Goh V. Rajan D.S. Tan A.D. Jeyasekharan Department of Haematology-Oncology National University Cancer Institute Singapore Singapore Department of Information Systems and Analytics School of Computing NUS - National University of Singapore Singapore Genomics and Data Analytics Cancer Science Institute (CSI) - National University of Singapore (NUS) Singapore National University Singapore Cancer Science Institute (CSI) - National University of Singapore (NUS) Singapore Medical Oncology Department NUHS - National University Health System Singapore

Time-varying formation tracking control for UAV-UGV heterogeneous swarm systems with switching directed topologies

学校读者我要写书评

暂无评论

Time-varying formation tracking control for UAV-UGV heteroge...

International Conference on Control and Automation (ICCA)

作者： Siquan Zhou Xiwang Dong Qingdong Li Zhang Ren School of Automation Science and Electrical Engineering Science and Technology on Aircraft Control Laboratory Beihang University Beijing P.R. China Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University Beijing P.R. China

ISBN: (数字)9781728190938

ISBN: (纸本)9781728190945

This paper studies time-varying formation tracking control problems for UAV-UGV heterogeneous swarm systems. The UAVs and UGVs need to achieve the desired time-varying formation and track the reference trajectory of a leader synchronously under switching directed topologies. Firstly, the cooperative control model for UAV-UGV heterogeneous swarm system is established by using algebraic graph theory. Then, the formation tracking control protocol based on a distributed observer is designed and the convergence is proved by using piecewise Lyapunov theory. Furthermore, the feasible conditions to achieve the time-varying formation tracking are analyzed, and the algorithm to select the control parameters is proposed. Finally, a numerical simulation example is provided to demonstrate the effectiveness of the proposed control approach.

关键词： Topology Switches Time-varying systems Observers Radar tracking Manganese Control systems

Non-invasive biopsy diagnosis of diabetic kidney disease via deep learning applied to retinal images: a population-based study

学校读者我要写书评

暂无评论

The Lancet Digital Health 2025年第5期7卷 100868页

作者： Meng, Ziyao Guan, Zhouyu Yu, Shujie Wu, Yilan Zhao, Yaoning Shen, Jie Lim, Cynthia Ciwei Chen, Tingli Yang, Dawei Ran, An Ran He, Feng Hamzah, Haslina Singh, Sarkaaj Abd Raof, Anis Syazwani Lee-Boey, Jian Wen Samuel Lim, Soo-Kun Sun, Xufang Ge, Shuwang Xu, Gang Su, Hua Cheng, Yang Lu, Feng Liao, Xiaofei Jin, Hai Deng, Chenxin Ruan, Lei Zhang, Cuntai Wu, Chan Dai, Rongping Jin, Yixiao Wang, Wenxiao Li, Tingyao Liu, Ruhan Li, Jiajia Shu, Jia Lu, Yuwei Wang, Xiangning Wu, Qiang Qin, Yiming Tang, Jin Sheng, Xiaohua Jiao, Qiong Yang, Xiaokang Guo, Minyi McKay, Gareth J Hogg, Ruth E Liew, Gerald Chee, Evelyn Yi Lyn Hsu, Wynne Lee, Mong Li Szeto, Simon Luk, Andrea O Y Chan, Juliana C N Cheung, Carol Y Tan, Gavin Siew Wei Tham, Yih-Chung Cheng, Ching-Yu Sabanayagam, Charumathi Lim, Lee-Ling Jia, Weiping Li, Huating Sheng, Bin Wong, Tien Yin Shanghai Belt and Road International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases Department of Computer Science and Engineering School of Electronic Information and Electrical Engineering Institute for Proactive Healthcare Shanghai Jiao Tong University Department of Endocrinology and Metabolism Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine Shanghai Diabetes Institute Shanghai Clinical Centre for Diabetes Shanghai Key Laboratory of Diabetes Mellitus Shanghai China MOE Key Laboratory of AI School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Shanghai China Beijing Tsinghua Changgung Hospital Eye Center School of Clinical Medicine Tsinghua Medicine Tsinghua University Beijing China Medical Records and Statistics Office Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine Shanghai China Department of Renal Medicine Singapore General Hospital SingHealth-Duke Academic Medical Centre Singapore Singapore Department of Ophthalmology Shanghai Health and Medical Center Wuxi China Department of Ophthalmology and Visual Sciences The Chinese University of Hong Kong Hong Kong Special Administrative Region China Singapore Eye Research Institute Singapore National Eye Centre Singapore Department of Medicine Faculty of Medicine Universiti Malaya Kuala Lumpur Malaysia Department of Ophthalmology Tongji Hospital Tongji Medical College Huazhong University of Science and Technology Hubei Wuhan China Department of Nephrology Tongji Hospital Tongji Medical College Huazhong University of Science and Technology Wuhan China Department of Nephrology Union Hospital Tongji Medical College Huazhong University of Science and Technology Wuhan China Department of Ophthalmology Union Hospital Tongji Medical College Huazhong University of Science and Technology Wuhan China National Engineering Research Centre for Big Dat

Background: Improving the accessibility of screening diabetic kidney disease (DKD) and differentiating isolated diabetic nephropathy from non-diabetic kidney disease (NDKD) are two major challenges in the field of diabetes care. We aimed to develop and validate an artificial intelligence (AI) deep learning system to detect DKD and isolated diabetic nephropathy from retinal fundus images. Methods: In this population-based study, we developed a retinal image-based AI-deep learning system, DeepDKD, pretrained using 734 084 retinal fundus images. First, for DKD detection, we used 486 312 retinal images from 121 578 participants in the Shanghai Integrated Diabetes Prevention and Care System for development and internal validation, and ten multi-ethnic datasets from China, Singapore, Malaysia, Australia, and the UK (65 406 participants) for external validation. Second, to differentiate isolated diabetic nephropathy from NDKD, we used 1068 retinal images from 267 participants for development and internal validation, and three multi-ethnic datasets from China, Malaysia, and the UK (244 participants) for external validation. Finally, we conducted two proof-of-concept studies: a prospective real-world study with 3 months' follow-up to evaluate the effectiveness of DeepDKD in screening DKD;and a longitudinal analysis of the effectiveness of DeepDKD in differentiating isolated diabetic nephropathy from NDKD on renal function changes with 4·6 years' follow-up. Findings: For detecting DKD, DeepDKD achieved an area under the receiver operating characteristic curve (AUC) of 0·842 (95% CI 0·838–0·846) on the internal validation dataset and AUCs of 0·791–0·826 across external validation datasets. For differentiating isolated diabetic nephropathy from NDKD, DeepDKD achieved an AUC of 0·906 (0·825–0·966) on the internal validation dataset and AUCs of 0·733–0·844 across external validation datasets. In the prospective study, compared with the metadata model, DeepDKD could detect DKD wit

关键词： Deep learning

Finite-Time Time-Varying Group Formation Tracking for High-Order Multi-Agent Systems with Mismatched Disturbances under Directed Graphs

学校读者我要写书评

暂无评论

Finite-Time Time-Varying Group Formation Tracking for High-O...

International Conference on Control and Automation (ICCA)

作者： Lei Tian Yongzhao Hua Xiwang Dong Qingdong Li Zhang Ren School of Automation Science and Electrical Engineering Science and Technology on Aircraft Control Laboratory Beihang University Beijing P.R. China Beijing Advanced Innovation Center for Big Data and Brain Computing Beihang University Beijing P.R. China

ISBN: (数字)9781728190938

ISBN: (纸本)9781728190945

Finite-time time-varying group formation (FT-TVGF) tracking control problems for high-order multi-agent systems (MASs) are studied in this paper. Different from previous work, the approaches proposed in this paper combine group formation tracking control with finite-time control, not only that, the agents in each group can achieve time-varying sub-formations and the relative configuration between different groups can be adjusted simultaneously. First, all agents are divided into three types, including the virtual leader, the group leader, and the follower. The virtual leader is added to plan the trajectory of MASs. Subgroups can be interacted with each other by the communication among group leaders. Followers achieve the finite-time formation tracking of group leaders. Then the disturbance observer is constructed to estimate the mismatched disturbances of all group leaders and followers in finite time. Furthermore, based on finite time control theory and observer theory, a distributed control protocol is presented to make MASs realize FT-TVGF tracking under directed graph. The proof of close-loop stability of MASs is accomplished using Lyapunov theory. Finally, numerical simulations verify the theoretical analysis of this paper.

关键词： Disturbance observers Trajectory Tracking Protocols Directed graphs Technological innovation Multi-agent systems