检索结果-内蒙古大学图书馆

Industrial Scene Text Detection with Refined Feature-attentive Network

学校读者我要写书评

暂无评论

arXiv 2021年

作者： Guan, Tongkun Gu, Chaochen Lu, Changsheng Tu, Jingzheng Feng, Qi Wu, Kaijie Guan, Xinping The Department of Automation Shanghai Jiao Tong University Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai Engineering Research Center of Intelligent Control and Management Shanghai200240 China The College of Engineering and Computer Science The Australian National University CanberraACT2600 Australia

Detecting the marking characters of industrial metal parts remains challenging due to low visual contrast, uneven illumination, corroded surfaces, and cluttered background of metal part images. Affected by these factors, bounding boxes generated by most existing methods could not locate low-contrast text areas very well. In this paper, we propose a refined feature-attentive network (RFN) to solve the inaccurate localization problem. Specifically, we first design a parallel feature integration mechanism to construct an adaptive feature representation from multi-resolution features, which enhances the perception of multi-scale texts at each scale-specific level to generate a high-quality attention map. Then, an attentive proposal refinement module is developed by the attention map to rectify the location deviation of candidate boxes. Besides, a re-scoring mechanism is designed to select text boxes with the best rectified location. To promote the research towards industrial scene text detection, we contribute two industrial scene text datasets, including a total of 102156 images and 1948809 text instances with various character structures and metal parts. Extensive experiments on our dataset and four public datasets demonstrate that our proposed method achieves the state-of-the-art performance. Both code and dataset are available at: https://***/TongkunGuan/RFN. © 2021, CC BY.

关键词： Character recognition

Double Backpropagation for Training Autoencoders against Adversarial Attack

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Sun, Chengjin Chen, Sizhe Huang, Xiaolin Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University MOE Key Laboratory of System Control and Information Processing 800 Dongchuan Road Shanghai200240 China

Deep learning, as widely known, is vulnerable to adversarial samples. This paper focuses on the adversarial attack on autoencoders. Safety of the autoencoders (AEs) is important because they are widely used as a compression scheme for data storage and transmission, however, the current autoencoders are easily attacked, i.e., one can slightly modify an input but has totally different codes. The vulnerability is rooted the sensitivity of the autoencoders and to enhance the robustness, we propose to adopt double backpropagation (DBP) to secure autoencoder such as VAE and DRAW. We restrict the gradient from the reconstruction image to the original one so that the autoencoder is not sensitive to trivial perturbation produced by the adversarial attack. After smoothing the gradient by DBP, we further smooth the label by Gaussian Mixture Model (GMM), aiming for accurate and robust classification. We demonstrate in MNIST, CelebA, SVHN that our method leads to a robust autoencoder resistant to attack and a robust classifier able for image transition and immune to adversarial attack if combined with GMM. Copyright © 2020, The Authors. All rights reserved.

关键词： Gaussian distribution

Distributed Frequency controller for MT-HVDC Systems Via Adaptive Dynamic Programming

学校读者我要写书评

暂无评论

Distributed Frequency Controller for MT-HVDC Systems Via Ada...

第32届中国控制与决策会议

作者： Zhongjie Hu Zhi-Wei Liu Xiong Hu Ming Chi College of Artificial Intelligence and Automation Huazhong University of Science and Technology the Key Laboratory of Image Processing and Intelligent Control Ministry of EducationHuazhong University of Science and Technology

The exact mathematical models of multi-terminal high-voltage direct current（MT-HVDC） systems are hard to be obtained in the practical application of HVDC *** overcome this challenge,a model-free distributed frequency controller for MT-HVDC systems is proposed based on Adaptive dynamic programming（ADP） in this ***,for each AC area,only the local and neighboring sampling data of frequency is required without the MT-HVDC system ***,the proposed controller is distributed which effectively balances the communication burden among the AC ***,the proposed controller makes the connected AC areas share their power reserves via HVDC grids to compensate load disturbances so that the necessary power reserves can be *** simulations carried out on a five-terminal HVDC system evaluate the performance of the proposed controller.

关键词： MT-HVDC system distributed frequency controller adaptive dynamic programming（ADP） neural network reinforcement learning

Adversarial semantic data augmentation for human pose estimation

学校读者我要写书评

暂无评论

arXiv 2020年

作者： Bin, Yanrui Cao, Xuan Chen, Xinya Ge, Yanhao Tai, Ying Wang, Chengjie Li, Jilin Huang, Feiyue Gao, Changxin Sang, Nong Key Laboratory of Image Processing and Intelligent Control School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China Tencent Youtu Lab China

Human pose estimation is the task of localizing body keypoints from still images. The state-of-the-art methods suffer from insufficient examples of challenging cases such as symmetric appearance, heavy occlusion and nearby person. To enlarge the amounts of challenging cases, previous methods augmented images by cropping and pasting image patches with weak semantics, which leads to unrealistic appearance and limited diversity. We instead propose Semantic Data Augmentation (SDA), a method that augments images by pasting segmented body parts with various semantic granularity. Furthermore, we propose Adversarial Semantic Data Augmentation (ASDA), which exploits a generative network to dynamiclly predict tailored pasting configuration. Given off-the-shelf pose estimation network as discriminator, the generator seeks the most confusing transformation to increase the loss of the discriminator while the discriminator takes the generated sample as input and learns from it. The whole pipeline is optimized in an adversarial manner. State-of-the-art results are achieved on challenging benchmarks. The code has been publicly available at https://***/Binyr/ASDA. Copyright © 2020, The Authors. All rights reserved.

关键词： Semantics

LinK3D: Linear keypoints Representation for 3D LiDAR Point Cloud

学校读者我要写书评

暂无评论

arXiv 2022年

作者： Cui, Yunge Zhang, Yinlong Dong, Jiahua Sun, Haibo Chen, Xieyuanli Zhu, Feng The Key Laboratory of Opto-Electronic Information Processing Shenyang Institute of Automation Chinese Academy of Sciences Shenyang China The Key Laboratory of Networked Control Systems Shenyang Institute of Automation Chinese Academy of Sciences Shenyang China The State Key Laboratory of Robotics Shenyang Institute of Automation Chinese Academy of Science Shenyang China The Institutes for Robotics and Intelligent Manufacturing Chinese Academy of Sciences Shenyang110016 China The University of Chinese Academy of Sciences Beijing101408 China The Shanghai Institute of Microsystem and Information Technology Chinese Academy of Sciences Shanghai China The College of Intelligence Science and Technology National University of Defense Technology Changsha China

Feature extraction and matching are the basic parts of many robotic vision tasks, such as 2D or 3D object detection, recognition, and registration. As is known, 2D feature extraction and matching have already achieved great success. Unfortunately, in the field of 3D, the current methods may fail to support the extensive application of 3D LiDAR sensors in robotic vision tasks due to their poor descriptiveness and inefficiency. To address this limitation, we propose a novel 3D feature representation method: Linear keypoints representation for 3D LiDAR point cloud, called LinK3D. The novelty of LinK3D lies in that it fully considers the characteristics (such as the sparsity and complexity) of LiDAR point clouds and represents the keypoint with its robust neighbor keypoints, which provide strong constraints in the description of the keypoint. The proposed LinK3D has been evaluated on three public datasets, and the experimental results show that our method achieves great matching performance. More importantly, LinK3D also shows excellent real-time performance, faster than the sensor frame rate at 10 Hz of a typical rotating LiDAR sensor. LinK3D only takes an average of 30 milliseconds to extract features from the point cloud collected by a 64-beam LiDAR and takes merely about 20 milliseconds to match two LiDAR scans when executed on a computer with an Intel Core i7 processor. Moreover, our method can be extended to LiDAR odometry task, and shows good scalability. We release the implementation of our method at https://***/YungeCui/LinK3D. Copyright © 2022, The Authors. All rights reserved.

关键词： Object detection

A novel method of detecting microangioma in retinal image based on visual characteristics 10

学校读者我要写书评

暂无评论

A novel method of detecting microangioma in retinal image ba...

10th International Conference on information Technology in Medicine and Education, ITME 2019

作者： Cao, Xinrong Lin, Hongkai Kang, Hongliang College of Computer and Control Engineering Minjiang University Fuzhou China Fujian Provincial Key Laboratory of Information Processing and Intelligent Control Minjiang University Fuzhou China

ISBN: (纸本)9781728139173

Diabetic retinopathy (DR), an ocular complication of diabetes could threaten vision, has become one of the common causes of blindness in adults' working age. The detection of microangioma (MA), the first feature and persists with the development of DR, is an important step in the diagnosis and follow-up of DR. However, due to complexity of retinal image and the randomness of abnormal lesions, the detection of MA is more susceptible to the influence of background texture noise and other structures in retinal image. A novel method for detecting MA based on visual features is proposed. It establishes an observation function and uses multi-range mean filtering to eliminate texture noise and vascular network. Then, the detection target is confirmed by the visual shape features. A series of tests in the different public retinal images databases show that the proposed method is effective and would be useful for computer-aided diagnosis of early screening of DR. © 2019 IEEE.

关键词： Computer aided diagnosis

Unsupervised Time Series Anomaly Detection under Data Contamination

学校读者我要写书评

暂无评论

Unsupervised Time Series Anomaly Detection under Data Contam...

IEEE International Symposium on information (IT) in Medicine and Education, ITME

作者： Xiaohui Lin Zuoyong Li Xunhua Huang Xinwei Chen Haoyi Fan College of Computer and Data Science Fuzhou University Fuzhou China Fujian Provincial Key Laboratory of Information Processing and Intelligent Control College of Computer and Control Engineering Minjiang University Fuzhou China School of Computer Science and Technology Harbin University of Science and Technology Harbin China Fujian Engineering and Research Center of New Chinese Lacquer Materials Minjiang University Fuzhou China School of Computer and Artificial Intelligence Zhengzhou University Zhengzhou China

Unsupervised learning utilizes unlabeled data to alleviate the reliance on large amounts of labeled data, and it has made great progress in time series anomaly detection. However, there are still some thorny issues unsolved, especially the incapability to maintain the robustness of the model when the time series dataset is contaminated. In this work, we propose an unsupervised time series anomaly detection method on data contamination, named distribution alignment autoencoder (DAAE), in which a temporal sliding reward function combines the adversarial distribution alignment to mine the contaminated data. Specifically, the dimensionality reduction and reconstruction of the time series are implemented by an autoencoder. Next, using two discriminators, the latent features obtained by dimensionality reduction are mapped to the Gaussian distribution, and the high-confidence normal time series are generated by sampling noise from the Gaussian distribution. Then, we compute the mean square error between high-confidence normal samples and the training data. Finally, we feed the error values to the temporal sliding reward function to mine the contaminated training samples. Experimental results compared with baselines on public datasets demonstrate the effectiveness of DAAE.

关键词： Training Dimensionality reduction Time series analysis Training data Mean square error methods Gaussian distribution Robustness

Research on the key Technology of Safety Protection in Application Environment of Blockchain 2

学校读者我要写书评

暂无评论

Research on the Key Technology of Safety Protection in Appli...

2019 2nd International Conference on Advanced Materials, intelligent Manufacturing and Automation, AMIMA 2019

作者： Zhang, Hua Wang, Jiong Xu, Ge Ruan, Zhiqiang College of Computer and Control Engineering Mmjiang University Fuzhou350108 China Fujian Provincial Key Laboratory of Information Processing and Intelligent Control Minjiang University China

This paper analyzes the current situation of blockchain application security. Then, makes a concrete analysis on the problems existing in the application safety of blockchain, including the attack of consensus algorithm, the leakage of privacy, and the programming vulnerability of blockchain and hash collisions. Then it introduces in detail the three key technologies of security protection under the blockchain application environment, P2P network technology, asymmetric encryption technology and consensus mechanism technology. In order to make more relevant people know more about the key technologies of safety protection in the application environment of blockchain. © Published under licence by IOP Publishing Ltd.

关键词： Cryptography

The Design of Logic Gate Based on Triplex Structures 14th

学校读者我要写书评

暂无评论

The Design of Logic Gate Based on Triplex Structures

14th International Conference on Bio-inspired Computing: Theories and Applications, BIC-TA 2019

作者： Hu, Yingxin Wang, Zhiyu Chen, Zhekun Liao, Sulin Huang, Yabing Key Laboratory of Image Information Processing and Intelligent Control of Education Ministry of China School of Artificial Intelligence and Automation Huazhong University of Science and Technology WuhanHubei430074 China College of Information Science and Technology Shijiazhuang Tiedao University ShijiazhuangHebei050043 China School of Computer Science and Technology Huazhong University of Science and Technology WuhanHubei430074 China Department of Pathology Renmin Hospital of Wuhan University WuhanHubei430060 China

ISBN: (纸本)9789811534140

Molecular carch hotspots in the area of DNA nanotechnology due to its inherent large storage capacity, high parallelism, and energy efficiency. The assembled structures in the existing computing units mainly adopt a double-helix structure. However, besides duplex, DNA is known to adopt other helical forms such as triplex form. The triplex structure has introduced a novel paradigm into the DNA engineer owing to their unique characteristic sequence, structural and assembly requirements. Based on the assembly and disassembly of the triplex structure, a YES logic gate was constructed via strand displacement reaction. The experiment results showed that GC content and stem length would influence the displacement reaction. Then AND logic gate was also established through cooperative effect by two DNA strands. The construction of the cascading circuit indicated triplex-based logic gate could be integrated into larger circuits and enrich the toolbox of the computing devices. © 2020, Springer Nature Singapore Pte Ltd.

关键词： Logic gates