检索结果-内蒙古大学图书馆

Overcoming Classifier Imbalance for Long-tail Object Detection with Balanced Group Softmax

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Li, Yu Wang, Tao Kang, Bingyi Tang, Sheng Wang, Chunfeng Li, Jintao Feng, Jiashi Key Laboratory of Intelligent Information Processing Institute of Computing Technology Chinese Academy of Sciences Beijing China University of Chinese Academy of Sciences Beijing China Department of Electrical and Computer Engineering National University of Singapore Singapore Institute of Data Science National University of Singapore Singapore

Solving long-tail large vocabulary object detection with deep learning based models is a challenging and demanding task, which is however under-explored. In this work, we provide the first systematic analysis on the underperformance of state-of-the-art models in front of long-tail distribution. We find existing detection methods are unable to model few-shot classes when the dataset is extremely skewed, which can result in classifier imbalance in terms of parameter magnitude. Directly adapting long-tail classification models to detection frameworks can not solve this problem due to the intrinsic difference between detection and classification. In this work, we propose a novel balanced group softmax (BAGS) module for balancing the classifiers within the detection frameworks through group-wise training. It implicitly modulates the training process for the head and tail classes and ensures they are both sufficiently trained, without requiring any extra sampling for the instances from the tail classes. Extensive experiments on the very recent long-tail large vocabulary object recognition benchmark LVIS show that our proposed BAGS significantly improves the performance of detectors with various backbones and frameworks on both object detection and instance segmentation. It beats all state-of-the-art methods transferred from long-tail image classification and establishes new state-of-the-art. Code is available at https://***/FishYuLi/ BalancedGroupSoftmax. Copyright © 2020, The Authors. All rights reserved.

关键词： Object recognition

Bilinear Semi-Tensor Product Attention (BSTPA) model for visual question answering

学校读者我要写书评

暂无评论

Bilinear Semi-Tensor Product Attention (BSTPA) model for vis...

International Joint Conference on Neural Networks (IJCNN)

作者： Zongwen Bai Ying Li Meili Zhou Di Li Dong Wang Dawid Połap Marcin Woźniak School of Computer Science Northwestern Polytechnical University Xi’an CHINA Shaanxi Key Laboratory of Intelligent Processing for Big Energy Data Yan’an CHINA School of Physics and Electronic Information Yan’an University Yan’an CHINA Faculty of Applied Mathematics Silesian University of Technology Gliwice POLAND

ISBN: (数字)9781728169262

ISBN: (纸本)9781728169279

We propose a semi-tensor product attention network model as a visual question answering tool for complex interaction over image features. Proposed model performs matrix multiplication of two arbitrary dimensions, which is used to overcome possible dimensional limitations and improve recognition flexibility. In used block-wise operation we preserve spatial and temporal information but reduce the number of parameters by using low-rank pooling scheme. Applied BERT pre-train model is tuned to recognize question features. The proposed model is evaluated on the VQA2.0 dataset. Research results show that our model has good accuracy and easy reconfiguration for future research.

关键词： Feature extraction Visualization Knowledge discovery Cognition Task analysis Computational modeling Bit error rate

Adaptive Hybrid Projective Models for Aerial Video Stitching

学校读者我要写书评

暂无评论

Adaptive Hybrid Projective Models for Aerial Video Stitching

International Conference on Systems and Informatics (ICSAI)

作者： Runping Xi Tianyi Han Yaqing Zhang Qianqian Han School of Computer Science Northwestern Polytechnical University National Engineering Laboratory of Integrated Aero-Space-Ground-Ocean Big Data Application Technology Shaanxi Provincial Key Laboratory of Speech & Image Information Processing Xi'an China

ISBN: (数字)9781728152561

ISBN: (纸本)9781728152578

Aerial video stitching has been widely used in many fields such as large-scale videos surveillance, disaster monitoring and unmanned aerial vehicle (UAV) navigation. Ghosting and parallax are challenging problems when the video streams are captured by UAV. This paper presents a novel method based on hybrid projective model which combines local warp alignment and global warp projection for better registration accuracy. The geometric alignment model is used for global registration of images and the content-preserving alignment model is used for local refinement. Simultaneously, reference frame is updated adaptively according to the shift of scenery, which is used to eliminate the cumulative error appearing in long video streams. The proposed approach has been tested under multiple datasets in different scenarios, which shows good performance.

关键词： Streaming media Feature extraction Adaptation models Cameras Image fusion Distortion Unmanned aerial vehicles

A Data Clustering Strategy for Enhancing Mutual Privacy in Healthcare System of IoT

学校读者我要写书评

暂无评论

A Data Clustering Strategy for Enhancing Mutual Privacy in H...

IEEE International Symposium on Parallel and Distributed processing with Applications and IEEE International Conference on Ubiquitous Computing and Communications (ISPA/IUCC)

作者： Xuancheng Guo Hui Lin Chuanfeng Xu Wenzhong Lin School of Mathematics and Computer Science Fujian Normal University Fuzhou China Fujian Provincial Key Laboratory of Information Processing and Intelligent control Minjiang University Fuzhou China

ISBN: (纸本)9781728152103

Recent advances in the healthcare system of Internet of Things (IoT) has led to a generation of a large amount of physic sensor data. Data analyst collects and analyzes these sensor data through wireless sensor network, so as to provide some treatment advices to physicians and patients. As a common data mining method, the k-means clustering algorithm is being applied to process large-scale sensor data. However, it also poses a threat of privacy leakage in the specific application process. To enhance the privacy in healthcare system of IoT, mutual privacypreserving k-means strategy (M-PPKS) based on homomorphic encryption is proposed in this paper, which neither discloses an individual's private information nor leaks the cluster center's characteristic data. An extension performance evaluation shows that, in the case of ensuring accurate clustering results, even if the analyst and individuals collude, the M-PPKS can prevent the disclosure of private information.

关键词： biomedical communication computer network security data analysis data mining data privacy health care Internet of Things medical computing pattern clustering wireless sensor networks Health Care Systems biomedical communication Network security Internet of things medical computing Data privacy Wireless sensor networks Cluster Analysis data mining Sensor Data Data analysis Health Care Privacy Homomorphic encryption

Learning a Layout Transfer Network for Context Aware Object Detection

学校读者我要写书评

暂无评论

arXiv 2019年

作者： Wang, Tao He, Xuming Cai, Yuanzheng Xiao, Guobao Fujian Provincial Key Laboratory of Information Processing and Intelligent Control Minjiang University Fuzhou China School of Information Science and Technology ShanghaiTech University Shanghai China College of Mathematics and Computer Science Fuzhou University NetDragon Fuzhou China

We present a context aware object detection method based on a retrieve-and-transform scene layout model. Given an input image, our approach first retrieves a coarse scene layout from a codebook of typical layout templates. In order to handle large layout variations, we use a variant of the spatial transformer network [1] to transform and refine the retrieved layout, resulting in a set of interpretable and semantically meaningful feature maps of object locations and scales. The above steps are implemented as a Layout Transfer Network which we integrate into Faster RCNN [2] to allow for joint reasoning of object detection and scene layout estimation. Extensive experiments on three public datasets verified that our approach provides consistent performance improvements to the state-of-the-art object detection baselines on a variety of challenging tasks in the traffic surveillance and the autonomous driving domains. Copyright © 2019, The Authors. All rights reserved.

关键词： Object detection

ICare Designer: A rule-driven layout co-designing system for elderly caring 23

学校读者我要写书评

暂无评论

ICare Designer: A rule-driven layout co-designing system for...

23rd IEEE International Conference on computer Supported Cooperative Work in Design, CSCWD 2019

作者： Sun, Yuling Zhou, Jie Yao, Dong He, Liang Department of Computer Science and Technology East China Normal University Shanghai China Shanghai Key Laboratory of Multidimensional Information Processing East China Normal University Shanghai China Izhaohu Co. Ltd. Shanghai China Department of Architecture Tongji University Shanghai China

ISBN: (纸本)9781728103501

In CSCW and related fields, computer supported collaborative design has been used in various fields. Yet, there has been little research focusing on the collaborative context design for elderly caring. For filling this gap, this paper explores a notion of rule-driven layout co-designing system for elderly caring. What makes it distinct from other layout designing system is that it employs hierarchically modularization to store all standardized components that are needed for the aging caring space, and embeds intelligence algorithm into the typical evidence-based design for improving design efficiency and reducing design costs. In this paper, we designed and implemented a prototype called iCare Designer, and conducted a preliminary study to understand how this notion works in reality. While findings of the preliminary study suggest promises of the notion of rule-driven space layout co-designing system, it reveals some challenges this kind of systems to truly work in practices, mainly reflecting in its flexibility and perception to ambient environment. An in-depth discussion is provided in the end. © 2019 IEEE.

关键词： Modular construction

Magnetotelluric signal-noise separation method based on SVM–CEEMDWT

学校读者我要写书评

暂无评论

Applied Geophysics 2019年第2期16卷 160-170,252-253页

作者： Li Jin Cai Jin Tang Jing-Tian Li Guang Zhang Xian Xu Zhi-Min Hunan Provincial Key Laboratory of Intelligent Computing and Language Information Processing College of Information Science and EngineeringHunan Normal UniversityChangsha 410081China School of Geosciences and Info-Physics Key Laboratory of Metallogenic Prediction of Non-Ferrous Metals and Geological Environment MonitorMinistry of EducationCentral South UniversityChangsha 410083China State Key Laboratory Breeding Base of Nuclear Resources and Environment East China University of TechnologyNanchang 330013China

To better retain useful weak low-frequency magnetotelluric(MT)signals with strong interference during MT data processing,we propose a SVM-CEEMDWT based MT data signal-noise separation method,which extracts the weak MT signal affected by strong ***,the approximate entropy,fuzzy entropy,sample entropy,and Lempel-Ziv(LZ)complexity are extracted from the magnetotelluric ***,four robust parameters are used as the inputs to the support vector machine(SVM)to train the sample library and build a model based on the different complexity of *** on this model,we can only consider time series with strong interference when using the complementary ensemble empirical mode decomposition(CEEMD)and wavelet threshold(WT)for noise *** results suggest that the SVM based on the robust parameters can distinguish the time periods with strong interference well before noise *** with the CEEMD WT,the proposed SVM-CEEMDWT method retains more low-frequency low-variability information,and the apparent resistivity curve is smoother and more ***,the results better reflect the deep electrical structure in the field.

关键词： SVM-CEEMDWT magnetotelluric signal-noise separation MT data processing

Low-latency federated learning over wireless channels with differential privacy

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Wei, Kang Li, Jun Ma, Chuan Ding, Ming Chen, Cailian Jin, Shi Han, Zhu Poor, H. Vincent School of Electrical and Optical Engineering Nanjing University of Science and Technology Nanjing210094 China Data61 CSIRO SydneyNSW2015 Australia The Department of Automation Shanghai Jiao Tong University Shanghai200240 China Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai200240 China The National Mobile Communications Research Laboratory Southeast University Nanjing210096 China The Department of Electrical and Computer Engineering The University of Houston HoustonTX77004 United States The Department of Computer Science and Engineering Kyung Hee University Seoul Korea Republic of The Department of Electrical and Computer Engineering Princeton University PrincetonNJ08544 United States

In federated learning (FL), model training is distributed over clients and local models are aggregated by a central server. The performance of uploaded models in such situations can vary widely due to imbalanced data distributions, potential demands on privacy protections, and quality of transmissions. In this paper, we aim to minimize FL training delay over wireless channels, constrained by overall training performance as well as each client's differential privacy (DP) requirement. We solve this problem in the framework of multi-agent multiarmed bandit (MAMAB) to deal with the situation where there are multiple clients confornting different unknown transmission environments, e.g., channel fading and interferences. Specifically, we first transform the long-term constraints on both training performance and each client's DP into a virtual queue based on the Lyapunov drift technique. Then, we convert the MAMAB to a max-min bipartite matching problem at each communication round, by estimating rewards with the upper confidence bound (UCB) approach. More importantly, we propose two efficient solutions to this matching problem, i.e., modified Hungarian algorithm and greedy matching with a better alternative (GMBA), in which the first one can achieve the optimal solution with a high complexity while the second one approaches a better tradeoff by enabling a verified low-complexity with little performance loss. In addition, we develop an upper bound on the expected regret of this MAMAB based FL framework, which shows a linear growth over the logarithm of communication rounds, justifying its theoretical feasibility. Extensive experimental results are conducted to validate the effectiveness of our proposed algorithms, and the impacts of various parameters on the FL performance over wireless edge networks are also discussed. Copyright © 2021, The Authors. All rights reserved.

关键词： Multi agent systems

Anchor-based spatio-temporal attention 3D convolutional networks for dynamic 3D point cloud sequences

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Wang, Guangming Chen, Muyao Liu, Hanwen Yang, Yehui Liu, Zhe Wang, Hesheng The Department of Automation Institute of Medical Robotics Key Laboratory of System Control and Information Processing Ministry of Education Key Laboratory of Marine Intelligent Equipment System of Ministry of Education Shanghai Jiao Tong University Shanghai200240 China Beijing Advanced Innovation Center for Intelligent Robots and Systems Beijing Institute of Technology China The Department of Computer Science and Engineering Shanghai Jiao Tong University Shanghai200240 China The Department of Computer Science and Technology University of Cambridge

With the rapid development of measurwement technology, LiDAR and depth cameras are widely used in the perception of the 3D environment. Recent learning based methods for robot perception most focus on the image or video, but deep learning methods for dynamic 3D point cloud sequences are underexplored. Therefore, developing efficient and accurate perception method compatible with these advanced instruments is pivotal to autonomous driving and service robots. An Anchor-based Spatio-Temporal Attention 3D Convolution operation (ASTA3DConv) is proposed in this paper to process dynamic 3D point cloud sequences. The proposed convolution operation builds a regular receptive field around each point by setting several virtual anchors around each point. The features of neighborhood points are firstly aggregated to each anchor based on the spatio-temporal attention mechanism. Then, anchor-based 3D convolution is adopted to aggregate these anchors' features to the core points. The proposed method makes better use of the structured information within the local region and learns spatio-temporal embedding features from dynamic 3D point cloud sequences. Anchor-based Spatio-Temporal Attention 3D Convolutional Neural Networks (ASTA3DCNNs) are built for classification and segmentation tasks based on the proposed ASTA3DConv and evaluated on action recognition and semantic segmentation tasks. The experiments and ablation studies on MSRAction3D and Synthia datasets demonstrate the superior performance and effectiveness of our method for dynamic 3D point cloud sequences. Our method achieves the state-of-the-art performance among the methods with dynamic 3D point cloud sequences as input on MSRAction3D and Synthia datasets. Copyright © 2020, The Authors. All rights reserved.

关键词： Embeddings