模式识别与人工智能
Home      About Journal      Editorial Board      Instructions      Ethics Statement      Contact Us                   中文
Pattern Recognition and Artificial Intelligence
22 Judgement and Disposal of Academic Misconduct Article
22 Copyright Transfer Agreement
22 Proof of Confidentiality
22 Requirements for Electronic Version
More....
22 Chinese Association of Automation
22 National ResearchCenter for Intelligent Computing System
22 Institute of Intelligent Machines,Chinese Academy of Sciences
More....
 
 
2020 Vol.33 Issue.11, Published 2020-11-25
Papers and Reports    Researches and Applications   
   
Papers and Reports
959 Gradient Controlled and Discriminative Features Guided Image-to-Image Translation Method Towards Realistic Portrait Illustrations
SHI Rongxiao, YE Dongyi, CHEN Zhaojiong

Without considering the preservation of facial recognition characteristics,the existing unsupervised image-to-image translation methods often suffer from face distortion and facial structure collapse.Consequently,it is difficult to identify the personal information after translation.To tackle this issue,gradient controlled and discriminative features guided image-to-image translation method towards realistic portrait illustrations is proposed based on cycle-consistent generative adversarial network(CycleGAN).Masked residual long connections are introduced by the proposed method to avoid reusing redundant features and image gradient information consistency is considered as a constraint to preserve facial recognition characteristics.In addition,a discriminative feature guided information-shared training mechanism is devised and thus generators are capable of capturing discriminative features of target images,like the discriminators.Moreover,patch-level discriminators are extended to multi-awareness discriminators to obtain more discriminative information.Experimental results show that the proposed method preserves facial recognition characteristics well in the translated illustrations and outperforms the existing unsupervised image-to-image translation methods in visual effect of illustrations.
2020 Vol. 33 (11): 959-971 [Abstract] ( 726 ) [HTML 1KB] [ PDF 5332KB] ( 576 )
972 Image Super-Resolution Reconstruction Based on Recursive Multi-scale Convolutional Networks
GAO Qingqing, ZHAO Jianwei, ZHOU Zhenghua
The performance of image super-resolution reconstruction networks is improved by deepening the depth.However,deepening the network makes the number of parameters increase rapidly,and thus it is hard to train the network and store the memory.To reduce the scale of the deep network and keep its reconstruction performance as much as possible,a concise recursive multi-scale convolutional network is proposed for super-resolution reconstruction based on the concepts of recursion and multi-scale.Firstly,the multi-scale module is employed to extract the features of the image with different scales.Then,the network is deepened by the recursive operation without increasing the number of network parameters.Finally,the outputs of each recursive operation are fused as the input for the reconstruction part.Experimental results show that the network parameters of the proposed method are fewer than those of some existing super-resolution methods with better reconstruction results.
2020 Vol. 33 (11): 972-980 [Abstract] ( 692 ) [HTML 1KB] [ PDF 2434KB] ( 724 )
981 Feature Selection Method Based on Improved Monarch Butterfly Optimization Algorithm
SUN Lin, ZHAO Jing, XU Jiucheng, XUE Zhan'ao
Aiming at the weak global search ability and the reduction of population diversity during migration of monarch butterfly optimization(MBO) algorithm,a differential adaptive MBO algorithm based on Cauchy mutation and its feature selection method are proposed.Firstly,the MBO migration operator is replaced by the mutation operation in the differential evolution algorithm to improve the global search ability.Then,MBO adjustment operator is combined with the adaptive adjustment strategy to change the single adjustment mode.Finally,Cauchy mutation is conducted in each updated population to increase population diversity.To verify the performance of the improved MBO algorithm and its feature selection method,experiments on benchmark functions and UCI datasets are conducted,and the results show that the proposed algorithms produce better performance than other algorithms.
2020 Vol. 33 (11): 981-994 [Abstract] ( 924 ) [HTML 1KB] [ PDF 977KB] ( 607 )
995 Interpretability Document Classification Based on Generative-Discriminatory Hybrid Model
WANG Qiang, CHEN Zhihao, XU Qing, BAO Liang, LIAO Xiangwen
The deep mining of text information and the semantic relationships between word and word context and between sentence and sentence context are not taken into account in the existing interpretability document classification.Therefore,an interpretable document classification method based on a generative-discriminatory-hybrid model is proposed.The hierarchical attention mechanism is introduced into the document encoder to obtain the document representations rich in contextual semantic information.And thus more accurate classification results and explanatory information are generated,and the problem of insufficient text information mining in the existing models is handled.Experiments on PCMag and Skytrax comment datasets show that the proposed method has better performance in document classification,generates more accurate explanatory information and improves the overall performance of the method.
2020 Vol. 33 (11): 995-1003 [Abstract] ( 422 ) [HTML 1KB] [ PDF 785KB] ( 335 )
1004 Vision Based Important Change Detection Method for Web Pages
SHI Cunhui, YU Xiaoming, LIU Yue, JIN Xiaolong, CHENG Xueqi
Duplicate Web indexes of Web crawling can be reduced effectively by detecting important changes and determining changes of essential content in Web pages.Therefore,a vision based detection method is proposed to detect changes in different semantic regions of the page and compress the page into a low dimensional vector representation.The proposed method is utilized to understand the difference of semantic importance in different regions from the perspective of users.Compared with the existing methods,the proposed method is independent of the analysis of HTML,and thus it is suitable for new media,such as mobile Internet.Experiments show the effectiveness of the proposed method.
2020 Vol. 33 (11): 1004-1012 [Abstract] ( 451 ) [HTML 1KB] [ PDF 1265KB] ( 665 )
Researches and Applications
1013 An Intelligent Fault Diagnosis Method Based on FastDTW for Railway Turnout
JI Wenjiang, ZUO Yuan, HEI Xinhong, SEI Takahashi, HIDEO Nakamura
The turnout handles the direction of the train.It is a key equipment for the safety of railway transportation system.An intelligent fault diagnosis method based on fast dynamic time warping(FastDTW) for railway turnout is proposed in this paper.It is testified by the real action current data obtained from switch machine model No.ZD7.Firstly,the original current curve is segmented according to wave form features.Then,the warp path distance between the standard sample and the tested current curve is obtained by FastDTW algorithm.Finally,a dynamic optimized threshold is exploited to confirm whether there is a fault in the turnout.The experimental results show the proposed method works well with both single and double action type turnout machines with only 200 turnout action current samples.The proposed method is suitable for the train control system of new generation as well due to its high diagnosis accuracy and low time cost.
2020 Vol. 33 (11): 1013-1022 [Abstract] ( 647 ) [HTML 1KB] [ PDF 888KB] ( 394 )
1023 Image Retrieval Based on Wavelet Projection and Discrete Hashing
RONG Mengjun, LIU Jinglei
Existing hashing methods can hardly realize the approximate mapping of the original feature space quickly.Therefore,a hashing method based on wavelet projection is proposed.Firstly,the projection matrix is constructed based on Haar wavelet transform.The projection matrix is optimized iteratively and binary codes are optimized by discrete method to control the quantization error.Then,the projection matrix is utilized to project the original feature vector of the image into the low-dimensional space quickly and binary codes are obtained by binary embedding.Experimental results on image datasets demonstrate that the proposed method improves encoding efficiency effectively.
2020 Vol. 33 (11): 1023-1032 [Abstract] ( 464 ) [HTML 1KB] [ PDF 692KB] ( 367 )
1033 Greedy Strategy Influence Maximization Algorithm Based on Seed Candidates
LI Meiling, QIAN Fulan, XU Tao, ZHAO Shu, ZHANG Yanping
The hill-climbing greedy algorithm is not easily extended to large-scale social networks due to its high time complexity.In this paper,it is theoretically analyzed that the node set influence evaluation can be transformed into local probability solution,and thus the algorithm efficiency is significantly improved.The local probability solution function is extend to the greedy algorithm.Based on seed candidates,the greedy influence maximization algorithm and the lazy forward influence maximization algorithm are proposed,respectively. Experiments on four real datasets show that the performance of the proposed algorithms is as high as that of cost-effective lazy forward selection,and the proposed algorithms are superior in running time.
2020 Vol. 33 (11): 1033-1042 [Abstract] ( 742 ) [HTML 1KB] [ PDF 904KB] ( 466 )
1043 Visual Question Answering Method Based on Yes/No Feedback
DENG Wei, WANG Jianming, JIN Guanghao
Aiming at the ambiguous question sentence in the visual question answering task,a visual question answering method based on Yes/No feedback is proposed.The Yes/No feedback mechanism is employed to determine whether or not the answer is correct for the first time.When the feedback given by the user is no,the question is re-analyzed,new questions are generated after disambiguation and different candidate answers are generated.The answer with the highest confidence is output as the final result.The experimental results on ClEVR,CLEVR-CoGen benchmark datasets show the proposed method achieves higher accuracy than the existing methods.
2020 Vol. 33 (11): 1043-1053 [Abstract] ( 417 ) [HTML 1KB] [ PDF 995KB] ( 278 )
模式识别与人工智能
 

Supervised by
China Association for Science and Technology
Sponsored by
Chinese Association of Automation
NationalResearchCenter for Intelligent Computing System
Institute of Intelligent Machines, Chinese Academy of Sciences
Published by
Science Press
 
Copyright © 2010 Editorial Office of Pattern Recognition and Artificial Intelligence
Address: No.350 Shushanhu Road, Hefei, Anhui Province, P.R. China Tel: 0551-65591176 Fax:0551-65591176 Email: bjb@iim.ac.cn
Supported by Beijing Magtech 　Email:support@magtech.com.cn