检索结果-内蒙古大学图书馆

MetaFuse: A Pre-trained Fusion Model for Human Pose Estimation

学校读者我要写书评

暂无评论

MetaFuse: A Pre-trained Fusion Model for Human Pose Estimati...

Conference on Computer Vision and Pattern Recognition (CVPR)

作者： Rongchang Xie Chunyu Wang Yizhou Wang Center for Data Science Peking University Deepwise AI Lab Microsoft Research Asia Adv. Inst. of Info. Tech. Peking University Center on Frontiers of Computing Studies Peking University CS Dept. Peking University

ISBN: (数字)9781728171685

ISBN: (纸本)9781728171692

Cross view feature fusion is the key to address the occlusion problem in human pose estimation. The current fusion methods need to train a separate model for every pair of cameras making them difficult to scale. In this work, we introduce MetaFuse, a pre-trained fusion model learned from a large number of cameras in the Panoptic dataset. The model can be efficiently adapted or finetuned for a new pair of cameras using a small number of labeled images. The strong adaptation power of MetaFuse is due in large part to the proposed factorization of the original fusion model into two parts-(1) a generic fusion model shared by all cameras, and (2) lightweight camera-dependent transformations. Furthermore, the generic model is learned from many cameras by a meta-learning style algorithm to maximize its adaptation capability to various camera poses. We observe in experiments that MetaFuse finetuned on the public datasets outperforms the state-of-the-arts by a large margin which validates its value in practice.

关键词： Cameras Adaptation models Three-dimensional displays Two dimensional displays Pose estimation Task analysis Heating systems

MetaFuse: A Pre-trained Fusion Model for Human Pose Estimation

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Xie, Rongchang Wang, Chunyu Wang, Yizhou Center for Data Science Peking University Adv. Inst. of Info. Tech. Peking University Center on Frontiers of Computing Studies Peking University CS Dept. Peking University Microsoft Research Asia Deepwise AI Lab

Cross view feature fusion is the key to address the occlusion problem in human pose estimation. The current fusion methods need to train a separate model for every pair of cameras making them difficult to scale. In this work, we introduce MetaFuse, a pre-trained fusion model learned from a large number of cameras in the Panoptic dataset. The model can be efficiently adapted or finetuned for a new pair of cameras using a small number of labeled images. The strong adaptation power of MetaFuse is due in large part to the proposed factorization of the original fusion model into two parts— (1) a generic fusion model shared by all cameras, and (2) lightweight camera-dependent transformations. Furthermore, the generic model is learned from many cameras by a meta-learning style algorithm to maximize its adaptation capability to various camera poses. We observe in experiments that MetaFuse finetuned on the public datasets outperforms the state-of-the-arts by a large margin which validates its value in practice. Copyright © 2020, The Authors. All rights reserved.

关键词： Cameras

Learning multi-agent coordination for enhancing target coverage in directional sensor networks 20

学校读者我要写书评

暂无评论

Learning multi-agent coordination for enhancing target cover...

Proceedings of the 34th International Conference on Neural information Processing Systems

作者： Jing Xu Fangwei Zhong Yizhou Wang Center for Data Science Peking University and Deepwise AI Lab Dept. of Computer Science Peking University and Adv. Inst. of Info. Tech Peking University and Advanced Innovation Center For Future Visual Entertainment Beijing Film Academy Dept. of Computer Science Peking University and Center on Frontiers of Computing Studies Peking University

ISBN: (纸本)9781713829546

Maximum target coverage by adjusting the orientation of distributed sensors is an important problem in directional sensor networks (DSNs). This problem is challenging as the targets usually move randomly but the coverage range of sensors is limited in angle and distance. Thus, it is required to coordinate sensors to get ideal target coverage with low power consumption, e.g. no missing targets or reducing redundant coverage. To realize this, we propose a Hierarchical Target-oriented Multi-Agent Coordination (HiT-MAC), which decomposes the target coverage problem into two-level tasks: targets assignment by a coordinator and tracking assigned targets by executors. Specifically, the coordinator periodically monitors the environment globally and allocates targets to each executor. In turn, the executor only needs to track its assigned targets. To effectively learn the HiT-MAC by reinforcement learning, we further introduce a bunch of practical methods, including a self-attention module, marginal contribution approximation for the coordinator, goal-conditional observation filter for the executor, etc. Empirical results demonstrate the advantage of HiT-MAC in coverage rate, learning efficiency, and scalability, comparing to baselines. We also conduct an ablative analysis on the effectiveness of the introduced components in the framework.

关键词：

A Cross-Virtual Machine Network Channel Attack via Mirroring and TAP Impersonation

学校读者我要写书评

暂无评论

A Cross-Virtual Machine Network Channel Attack via Mirroring...

IEEE International Conference on Cloud computing, CLOUD

作者： Atif Saeed Peter Garraghan Barnaby Craggs Dirk van der Linden Awais Rashid Syed Asad Hussain School of Computing and Communications Lancaster University Department of Computer Science University of Bristol Department of Computer Science COMSATS Inst. of Information Tech.

data privacy and security is a leading concern for providers and customers of cloud computing, where Virtual Machines (VMs) can co-reside within the same underlying physical machine. Side channel attacks within multi-tenant virtualized cloud environments are an established problem, where attackers are able to monitor and exfiltrate data from co-resident VMs. Virtualization services have attempted to mitigate such attacks by preventing VM-to-VM interference on shared hardware by providing logical resource isolation between co-located VMs via an internal virtual network. However, such approaches are also insecure, with attackers capable of performing network channel attacks which bypass mitigation strategies using vectors such as ARP Spoofing, TCP/IP steganography, and DNS poisoning. In this paper we identify a new vulnerability within the internal cloud virtual network, showing that through a combination of TAP impersonation and mirroring, a malicious VM can successfully redirect and monitor network traffic of VMs co-located within the same physical machine. We demonstrate the feasibility of this attack in a prominent cloud platform - OpenStack - under various security requirements and system conditions, and propose countermeasures for mitigation.

关键词： Bridges Cloud computing IP networks Security Virtual machine monitors Switches Monitoring

Optimal algorithm for profiling dynamic arrays with finite values

学校读者我要写书评

暂无评论

arXiv 2018年

作者： Yang, Dingcheng Yu, Wenjian Deng, Junhui Liu, Shenghua BNRist Dept. Computer Science & Tech. Tsinghua Univ. Beijing China CAS Key Lab. Network Data Science & Tech. Inst. Computing Technology Chinese Academy of Sciences Beijing China

How can one quickly answer the most and top popular objects at any time, given a large log stream in a system of billions of users? It is equivalent to find the mode and top-frequent elements in a dynamic array corresponding to the log stream. However, most existing work either restrain the dynamic array within a sliding window, or do not take advantages of only one element can be added or removed in a log stream. Therefore, we propose a profiling algorithm, named S-Profile, which is of O(1) time complexity for every updating of the dynamic array, and optimal in terms of computational complexity. With the profiling results, answering the queries on the statistics of dynamic array becomes trivial and fast. With the experiments of various settings of dynamic arrays, our accurate S-Profile algorithm outperforms the well-known methods, showing at least 2X speedup to the heap based approach and 13X or larger speedup to the balanced tree based approach. Copyright © 2018, The Authors. All rights reserved.

关键词： Computational complexity

The machine learning landscape of top taggers

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Kasieczka, G. Plehn, T. Butter, A. Cranmer, K. Debnath, D. Dillon, B.M. Fairbairn, M. Faroughy, D.A. Fedorko, W. Gay, C. Gouskos, L. Kamenik, J.F. Komiske, P.T. Leiss, S. Lister, A. Macaluso, S. Metodiev, E.M. Moore, L. Nachman, B. Nordström, K. Pearkes, J. Qu, H. Rath, Y. Rieger, M. Shih, D. Thompson, J.M. Varma, S. Institut für Experimentalphysik Universität Hamburg Germany Institut für Theoretische Physik Universität Heidelberg Germany Center for Cosmology and Particle Physics and Center for Data Science NYU United States NHECT Dept. of Physics and Astronomy Rutgers State University of NJ United States Jozef Stefan Institute Ljubljana Slovenia Theoretical Particle Physics and Cosmology King’s College London United Kingdom Department of Physics and Astronomy University of British Columbia Canada Department of Physics University of California Santa Barbara United States Faculty of Mathematics and Physics University of Ljubljana Ljubljana Slovenia Center for Theoretical Physics MIT Cambridge United States CP3 Universitéxx Catholique de Louvain Louvain-la-Neuve Belgium Physics Division Lawrence Berkeley National Laboratory Berkeley United States Simons Inst. for the Theory of Computing University of California Berkeley United States Amsterdam Netherlands LPTHE CNRS & Sorbonne Université Paris France III. Physics Institute A RWTH Aachen University Germany

Based on the established task of identifying boosted, hadronically decaying top quarks, we compare a wide range of modern machine learning approaches. Unlike most established methods they rely on low-level input, for inst.nce calorimeter output. While their network architectures are vastly different, their performance is comparatively similar. In general, we find that these new approaches are extremely powerful and great fun. Copyright © 2019, The Authors. All rights reserved.

关键词： Network architecture

Vibration isolation system with a compact damping system for power recycling mirrors of KAGRA

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Akiyama, Y. Akutsu, T. Ando, M. Arai, K. Arai, Y. Araki, S. Araya, A. Aritomi, N. Asada, H. Aso, Y. Bae, S. Baiotti, L. Barton, M.A. Cannon, K. Capocasa, E. Chen, C.-S. Chiu, T.-W. Cho, K. Chu, Y.-K. Craig, K. Dattilo, V. Doi, K. Enomoto, Y. Flaminio, R. Fujii, Y. Fujimoto, M.-K. Fukunaga, M. Fukushima, M. Furuhata, T. Haino, S. Hasegawa, K. Hashimoto, Y. Hashino, K. Hayama, K. Hirayama, T. Hirose, E. Hsieh, B.H. Huang, C.-Z. Ikenoue, B. Inoue, Y. Ioka, K. Itoh, Y. Izumi, K. Kaji, T. Kajita, T. Kakizaki, M. Kamiizumi, M. Kanbara, S. Kanda, N. Kanemura, S. Kang, G. Kasuya, J. Kawai, N. Kawasaki, T. Kim, C. Kim, W.S. Kim, J. Kim, J.C. Kimura, N. Kirii, S. Kitaoka, Y. Kitazawa, H. Kojima, Y. Kokeyama, K. Komori, K. Kong, A. Kotake, K. Kozu, R. Kumar, R. Kuo, H.-S. Kuroki, S. Kuroyanagi, S. Lee, H.K. Lee, H.M. Lee, H.W. Leonardi, M. Lin, C.-Y. Lin, F.-L. Liu, G.C. Marchio, M. Matsui, T. Michimura, Y. Mio, N. Miyakawa, O. Miyamoto, A. Miyoki, S. Morii, W. Morisaki, S. Moriwaki, Y. Musha, M. Nagano, S. Nagano, K. Nakamura, K. Nakamura, T. Nakano, H. Nakano, M. Narikawa, T. Quynh, L. Nguyen Ni, W.-T. Nishizawa, A. Obuchi, Y. Oh, J. Oh, S.H. Ohashi, M. Ohishi, N. Ohkawa, M. Okutomi, K. Ono, K. Oohara, K. Ooi, C.P. Pan, S.-S. Paoletti, F. Park, J. Passaquieti, R. Peña Arellano, F.E. Sago, N. Saito, S. Saito, Y. Sakai, K. Sakai, Y. Sasai, M. Sato, S. Sato, T. Sekiguchi, T. Sekiguchi, Y. Shibata, M. Shimoda, T. Shinkai, H. Shishido, T. Shoda, A. Someya, N. Somiya, K. Son, E.J. Suemasa, A. Suzuki, T. Suzuki, T. Tagoshi, H. Tahara, H. Takahashi, H. Takahashi, R. Takeda, H. Tanaka, H. Tanaka, K. Tanaka, T. Tanioka, S. San Martin, E.N. Tapia Tomaru, T. Tomura, T. Travasso, F. Tsubono, K. Tsuchida, S. Uchikata, N. Uchiyama, T. Uehara, T. Ueno, K. Uraguchi, F. Ushiba, T. van Putten, M.H.P.M. Vocca, H. Wakamatsu, T. Watanabe, Y. Xu, W.-R. Yamada, T. Yamamoto, K. Yamamoto, K. Yamamoto, S. Yamamoto, T. Yokogawa, K. Yokoyama, J. Yokozawa, T. Yoshioka, T. Yuzurihara, H. Zeidler, S. Zhu, Z.-H. Graduate School of Science and Engineering Hosei University TokyoKoganei184-8584 Japan National Astronomical Observatory of Japan TokyoMitaka181-8588 Japan Department of Physics University of Tokyo TokyoBunkyo113-0033 Japan University of Tokyo KashiwaChiba277-8582 Japan High Energy Accelerator Research Organization TsukubaIbaraki305-0801 Japan Earthquake Research Institute University of Tokyo TokyoBunkyo113-0032 Japan Department of Physics University of Tokyo TokyoBunkyo113-0033 Japan Department of Mathematics and Physics Hirosaki University AomoriHirosaki036-8561 Japan National Astronomical Observatory of Japan GifuHida506-1205 Japan Korea Institute of Science and Technology Information Yuseong Daejeon34141 Korea Republic of Department of Physics Osaka University ToyonakaOsaka560-0043 Japan University of Tokyo TokyoBunkyo113-0033 Japan Department of Physics National Taiwan Normal University Taipei116 Taiwan Department of Physics Sogang University Seoul121-742 Korea Republic of PisaCascinaI-56021 Italy Department of Physics University of Toyama ToyamaToyama930-8555 Japan Department of Astronomy University of Tokyo TokyoBunkyo113-0032 Japan Advanced Technology Center National Astronomical Observatory of Japan TokyoMitaka181-8588 Japan Institute of Physics Academia Sinica TaipeiNankang11529 Taiwan Department of Applied Physics Fukuoka University FukuokaJonan814-0180 Japan Yukawa Institute for Theoretical Physics Kyoto University KyotoSakyo606-8502 Japan Graduate School of Science Osaka City University OsakaSumiyosi558-8585 Japan Institute of Space and Astronautical Science Japan Aerospace Exploration Agency SagamiharaKanagawa252-5210 Japan University of Tokyo GifuHida506-1205 Japan Graduate Schoool of Science and Technology Tokyo Institute of Technology MeguroTokyo152-8551 Japan Department of Physics Ewha Womans University SeoulSeodaemun-gu03760 Korea Republic of National Institute for

A vibration isolation system called Type-Bp system used for power recycling mirrors has been developed for KAGRA, the interferometric gravitational-wave observatory in Japan. A suspension of the Type-Bp system passively isolates an optic from seismic vibration using three main pendulum stages equipped with two vertical vibration isolation systems. A compact reaction mass around each of the main stages allows for achieving sufficient damping performance with a simple feedback as well as vibration isolation ratio. Three Type-Bp systems were inst.lled in KAGRA, and were proved to satisfy the requirements on the damping performance, and also on estimated residual displacement of the optics. Copyright © 2019, The Authors. All rights reserved.

关键词： Mirrors

Exploring outliers in crowdsourced ranking for QoE

学校读者我要写书评

暂无评论

arXiv 2017年

作者： Xu, Qianqian Yan, Ming Huang, Chendi Xiong, Jiechao Huang, Qingming Yao, Yuan Institute of Information Engineering Cas Beijing100093 Department of Computational Mathematics Michigan State University East LansingMI48824 United States BICMR-LMAM-LMEQF-LMP School of Mathematical Sciences Peking University Beijing100871 Tencent Ai Lab Shenzhen518057 University of Chinese Academy of Sciences Beijing100049 China Key Lab of Intell. Info. Process. Inst. of Comput. Tech. Cas Beijing100190 Key Lab of Big Data Mining and Knowledge Management Cas Beijing100190 Department of Mathematics Hong Kong University of Science and Technology 100871 Hong Kong

Outlier detection is a crucial part of robust evaluation for crowd-sourceable assessment of Quality of Experience (QoE) and has attracted much attention in recent years. In this paper, we propose some simple and fast algorithms for outlier detection and robust QoE evaluation based on the nonconvex optimization principle. Several iterative procedures are designed with or without knowing the number of outliers in samples. Theoretical analysis is given to show that such procedures can reach statistically good estimates under mild conditions. Finally, experimental results with simulated and real-world crowdsourcing datasets show that the proposed algorithms could produce similar performance to Huber-LASSO approach in robust ranking, yet with nearly 8 or 90 times speed-up, without or with a prior knowledge on the sparsity size of outliers, respectively. Therefore the proposed methodology provides us a set of helpful tools for robust QoE evaluation with crowdsourcing data. Copyright © 2017, The Authors. All rights reserved.

关键词： Crowdsourcing

Predictive and preventive models for diabetes prevention using clinical information in electronic health record

学校读者我要写书评

暂无评论

Predictive and preventive models for diabetes prevention usi...

IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2015

作者： Cao, Ni Zeng, Sisi Shen, Feixia Pan, Chuandi Chen, Chengshui Nguyen, Thanh Chen, Jake Inst. of Biopharmaceutical Informatics and Technologies Wenzhou Medical University Zhejiang China First Affiliated Hospital Wenzhou Medical University Zhejiang China Dept. of Computer and Information Science Indiana University Purdue University IndianapolisIN United States School of Informatics and Computing Indiana University Purdue University IndianapolisIN United States

ISBN: (纸本)9781467367981

In this work, we constructed diabetes predictive models using electronic health record data, which could potentially have better preventive power than other diabetes predictive models known according to our knowledge. Diabetes is one of the most common, costly and complicated diseases all over the world, including China. To tackle the complexity of diabetes, electronic health record has been widely used to support physicians in integrated care. However, diabetes predictive models using electronic health record may lack of preventive power when the clinical measurements directly related to diabetes diagnosis criteria are used. To overcome this limitation, we did not use glucose, insulin, C-peptide and HbA1C clinical measurements in classifying diabetes patients. We used decision-table and support vector machine algorithm to build predictive models. As the result, our decision-table-based model achieves accuracy of 0.879, AUC of 0.921, precision of 0.898 and recall of 0.904, which is comparable with any known definition to diabetes. Our support-vector-machine-based model achieves accuracy of 0.660, AUC of 0.584, precision of 0.652 and recall of 0.939. We also found 37 measurements significantly associated to diabetes, which are not directly related to diabetes diagnosis criteria. © 2015 IEEE.

关键词： Support vector machines