Most Download articles

    Published in last 1 year | In last 2 years| In last 3 years| All| Most Downloaded in Recent Month | Most Downloaded in Recent Year|

    In last 2 years
    Please wait a minute...
    For Selected: Toggle Thumbnails
    Computer Engineering and Applications    2020, 56 (24): 0-0.  
    Abstract197)      PDF(pc) (1126KB)(1322)       Save
    Related Articles | Metrics
    Survey of Data Fusion Based on Deep Learning
    ZHANG Hong, CHENG Chuanqi, XU Zhigang, LI Jianhua
    Computer Engineering and Applications    2020, 56 (24): 1-11.   DOI: 10.3778/j.issn.1002-8331.2007-0475
    Abstract724)      PDF(pc) (683KB)(1045)       Save

    As data fusion is the key to maximize the value of big data, while deep learning is a technical tool for mining deep characteristic information of data, data fusion based on deep learning can fully tap the potential value of big data, thus expanding the exploration and understanding of the world to a new depth and breadth. And this paper learns the advantages of deep learning in data fusion by reviewing the literature related to data fusion based on deep learning in recent years. The common data-fusion methods are classified, the advantages and disadvantages of which are pointed out. Analysis is conducted on data fusion method based on deep learning from three perspectives, namely the data fusion method extracted based on features of deep learning, data fusion method based on deep learning fusion and data fusion method based on the whole process of deep learning, and corresponding comparisons and summaries are conducted as well. This paper summarizes the whole document, discusses the difficulties in the application of deep learning in data fusion and the problems which require further research in the future.

    Related Articles | Metrics
    Survey of Intelligent Question Answering Research Based on Knowledge Graph
    WANG Zhiyue, YU Qing, WANG Nan, WANG Yaoguo
    Computer Engineering and Applications    2020, 56 (23): 1-11.   DOI: 10.3778/j.issn.1002-8331.2004-0370
    Abstract785)      PDF(pc) (774KB)(1043)       Save

    The answer selection model based on knowledge graph has become one of the hottest directions at present. This paper introduces the implementation of answer selection model based on knowledge graph from four aspects of template method, semantic parsing, deep learning and knowledge graph embedding, sums up their advantages, disadvantages and unsolved problem. Combined with the development of artificial intelligence technology, this paper introduces intelligent question-answer system based on deep learning. This research is helpful for more researchers to devote themselves to the intelligent question-answer system and develops different kinds of intelligent question-answer system to improve the social intelligent information service.

    Related Articles | Metrics
    Review of Application Prospect of Deep Reinforcement Learning in Intelligent Manufacturing
    KONG Songtao, LIU Chichi, SHI Yong, XIE Yi, WANG Kun
    Computer Engineering and Applications    2021, 57 (2): 49-59.   DOI: 10.3778/j.issn.1002-8331.2008-0431
    Abstract601)      PDF(pc) (982KB)(1030)       Save

    As the latest development of machine learning, deep reinforcement learning has been shown in many application fields. The algorithm research and application research of deep reinforcement learning have produced many classical algorithms and typical application fields. The application of deep reinforcement learning in industrial manufacturing can realize high level control in complex environment. First of all, the research on deep reinforcement learning is summarized, and the basic principles of deep reinforcement learning are introduced, including deep learning and reinforcement learning. Then, the paper introduces the theoretical methods of the application of deep reinforcement learning algorithm. On this basis, it classifies the algorithms of deep reinforcement learning, respectively introduces the reinforcement learning algorithm based on value function and the reinforcement learning algorithm based on strategy gradient, and lists the main development results of these two kinds of algorithms, as well as other related research results. Then, the typical applications of deep reinforcement learning in industrial manufacturing are classified and analyzed. Finally, the existing problems and future development direction of deep reinforcement learning are discussed.

    Related Articles | Metrics
    Review of Research on Generative Adversarial Networks and Its Application
    WEI Fuqiang, Gulanbaier Tuerhong, Mairidan Wushouer
    Computer Engineering and Applications    2021, 57 (19): 18-31.   DOI: 10.3778/j.issn.1002-8331.2104-0248
    Abstract386)      PDF(pc) (1078KB)(994)       Save

    The theoretical research and applications of generative adversarial networks have been continuously successful and have become one of the current hot spots of research in the field of deep learning. This paper provides a systematic review of the theory of generative adversarial networks and their applications in terms of types of models, evaluation criteria and theoretical research progress; analyzing the strengths and weaknesses of generative models with explicit and implicit density-based, respectively; summarizing the evaluation criteria of generative adversarial networks, interpreting the relationship between the criteria, and introduces the research progress of the generative adversarial network in image generation from the application level, that is, through the image conversion, image generation, image restoration, video generation, text generation and image super-resolution applications; analyzing the theoretical research progress of generative adversarial networks from the perspectives of interpretability, controllability, stability and model evaluation methods. Finally, the paper discusses the challenges of studying generative adversarial networks and looks forward to the possible future directions of development.

    Reference | Related Articles | Metrics
    Survey of Image Compression Algorithm Based on Deep Learning
    YU Heng, MEI Hongyan, XU Xiaoming, JIA Huiping
    Computer Engineering and Applications    2020, 56 (15): 15-23.   DOI: 10.3778/j.issn.1002-8331.2003-0294
    Abstract481)      PDF(pc) (923KB)(939)       Save

    With the continuous development of deep learning and the explosive growth of image data, how to use deep learning to obtain higher compression ratio and higher quality images has gradually become one of the hot research issues. Through the analysis of the related literatures in recent years, the image compression method based on the deep learning is summarized and analyzed according to the Convolutional Neural Network(CNN), Recurrent Neural Network(RNN), Generative Adversarial Network(GAN). This paper enumerates the typical examples, and the image compression algorithm based on depth study of the training data set, commonly used evaluation indexes are introduced, according to the deep learning advantages in the field of image compression for its future development trend are summarized and discussed.

    Related Articles | Metrics
    Survey of Small Object Detection Algorithms Based on Deep Learning
    LIU Yang, ZHAN Yinwei
    Computer Engineering and Applications    2021, 57 (2): 37-48.   DOI: 10.3778/j.issn.1002-8331.2009-0047
    Abstract915)      PDF(pc) (959KB)(917)       Save

    With the development of artificial intelligence technology, deep learning technology has been widely used in face recognition, pedestrian detection, unmanned driving and other fields. As one of the most basic and challenging problems in machine vision, object detection has attracted extensive attention in recent years. Aiming at the problem of object detection, especially small object detection, this paper summarizes the common data sets and performance evaluation metrics, and compares the characteristics, advantages and difficulties of various common data sets. At the same time, this paper systematically summarizes the common object detection methods and the challenges faced by small object detection. In addition, combing the latest work based on deep learning, this paper introduces the multi-scale and super-resolution small object detection methods in the highlight and presents the lightweight strategy and the performance of some lightweight models based on the object detection. Finally, this paper summarizes the characteristics, advantages and limitations of various methods, and looks at the future development direction of small object detection method based on deep learning.

    Related Articles | Metrics
    3D Path Planning Algorithm Based on Deep Reinforcement Learning
    HUANG Dongjin, JIANG Chenfeng, HAN Kaili
    Computer Engineering and Applications    2020, 56 (15): 30-36.   DOI: 10.3778/j.issn.1002-8331.2001-0347
    Abstract422)      PDF(pc) (1152KB)(899)       Save

    Reasonable path selection is a difficulty in the field of 3D path planning. The existing 3D path planning methods can not adapt to the unknown terrain, and the obstacle avoidance form is single. In order to solve these problems, a 3D path planning algorithm for agents based on LSTM-PPO is proposed. Virtual ray is designed to detect simulation environment, and the collected state space and action states are introduced into Long Short-Term Memory Networks(LSTM). Through the extra reward function and intrinsic curiosity module, the agent can learn to jump through low obstacles and avoid large obstacles. Using the PPO’s clipped surrogate objective to optimize the update range of planning strategy. The results show that the algorithm is feasible, more intelligent and more reasonable for path planning, and can adapt well to the unknown environment with many obstacles.

    Related Articles | Metrics
    Review of Machine Learning for Predictive Maintenance
    LI Jieqi, HU Liangbing
    Computer Engineering and Applications    2020, 56 (21): 11-19.   DOI: 10.3778/j.issn.1002-8331.2006-0016
    Abstract338)      PDF(pc) (660KB)(883)       Save

    Machine learning algorithms can process high-dimensional and multi-variable data, and extract hidden relationships in the data in complex and dynamic environments, and have good application prospects in predictive maintenance technology. However, the performance of predictive maintenance system depends on the choice of machine learning algorithms. This paper reviews the current machine learning algorithms used in predictive maintenance system, compares the advantages and disadvantages of several machine learning algorithms characteristic in detail. The application of the machine learning in predictive maintenance is prospected in the future.

    Related Articles | Metrics
    Survey of Breast Cancer Histopathology Image Classification Based on Deep Learning
    LI Hua, YANG Jianeng, LIU Feng, NAN Fangzhe, QIAN Yurong
    Computer Engineering and Applications    2020, 56 (13): 1-11.   DOI: 10.3778/j.issn.1002-8331.2001-0220
    Abstract532)      PDF(pc) (919KB)(879)       Save

    Accurate and efficient histopathological image classification of breast cancer is one of the important contents of computer-aided diagnosis. With the development of machine learning technology, deep learning has gradually become an effective method to classify breast cancer histopathological images. Firstly, the classification methods of breast cancer histopathological image and the existing problems are analyzed. Secondly, four relevant deep learning models are introduced, and the classification methods of breast cancer histopathological image based on deep learning are combed, and the performance of the existing models is compared and analyzed through experiments. Finally, the key issues of histopathological image classification of breast cancer are summarized and the future research trends are discussed.

    Related Articles | Metrics
    Survey of Medical Image Segmentation Algorithm in Deep Learning
    PENG Jing, LUO Haoyu, ZHAO Gansen, LIN Chengchuang, YI Xusheng, CHEN Shaojie
    Computer Engineering and Applications    2021, 57 (3): 44-57.   DOI: 10.3778/j.issn.1002-8331.2010-0335
    Abstract436)      PDF(pc) (1397KB)(835)       Save

    Medical image segmentation is an important application area of computer vision in the medical image processing, its goal is to segment the target area from medical images and provide effective help for subsequent diagnosis and treatment of diseases. Since deep learning technology has made great progress in the image processing, medical image segmentation algorithm based on deep learning has gradually become the focus and hotspot of research in this field. This paper gives a description on the tasks and difficulties of medical image segmentation. Then, it details the deep learning-based medical image segmentation algorithm, classifies and summarizes the current representative methods. Moreover, this paper presents the frequently-used algorithm evaluation indicators and datasets in the field of medical image segmentation. The development of medical image segmentation technology is summarized and forecasted.

    Related Articles | Metrics
    Analysis on Application of Machine Learning in Stock Forecasting
    XU Haoran, XU Bo, XU Kewen
    Computer Engineering and Applications    2020, 56 (12): 19-24.   DOI: 10.3778/j.issn.1002-8331.2001-0353
    Abstract586)      PDF(pc) (892KB)(829)       Save

    It has always been regarded as the emphasis of research to reveal the operation law of stock market. In recent years, machine learning method has made good progress in stock forecasting, and it has shown unique advantages over traditional methods such as fundamental analysis and technical analysis. This paper focuses on collecting the key references in the field of stock prediction that uses machine learning methods in recent years, and analyzing as well as summarizing feature engineering, the application of machine learning algorithms and the main problems in stock prediction research. The characteristics and shortcomings of each algorithm in application are reviewed, and future development direction of this field is made a thorough analysis and forecasted from the aspects of transfer learning, feature engineering and deep learning model fusion.

    Related Articles | Metrics
    Survey of Network Traffic Forecast Based on Deep Learning
    KANG Mengxuan, SONG Junping, FAN Pengfei, GAO Bowen, ZHOU Xu, LI Zhuo
    Computer Engineering and Applications    2021, 57 (10): 1-9.   DOI: 10.3778/j.issn.1002-8331.2101-0402
    Abstract900)      PDF(pc) (711KB)(820)       Save

    Precisely predicting the trend of network traffic changes can help operators accurately predict network usage, correctly allocate and efficiently use network resources to meet the growing and diverse user needs. Taking the progress of deep learning algorithms in the field of network traffic prediction as a clue, this paper firstly elaborates the evaluation indicators of network traffic prediction and the current public network traffic data sets. Secondly, this paper specifically analyzes four deep learning methods commonly used in network traffic prediction:deep belief networks, convolutional neural network, recurrent neural network, and long short term memory network, and focuses on the integrated neural network models used in recent years for different problems. The characteristics and application scenarios of each model are summarized. Finally, the future development of network traffic forecast is prospected.

    Related Articles | Metrics
    Improved Algorithm of RetinaFace for Natural Scene Mask Wear Detection
    NIU Zuodong, QIN Tao, LI Handong, CHEN Jinjun
    Computer Engineering and Applications    2020, 56 (12): 1-7.   DOI: 10.3778/j.issn.1002-8331.2002-0402
    Abstract786)      PDF(pc) (1216KB)(816)       Save

    The 2019-nCoV can be transmitted through airborne droplets, aerosols, and other carriers. Correctly wearing a mask in public places can effectively prevent the infection of the virus. A face mask wearing detection method in a natural scene is proposed. The RetinaFace algorithm is improved to add the task of face mask wearing detection by optimizing the loss function. An improved self-attention mechanism is introduced into the feature pyramid network to enhance the expressive ability of the feature map. A data set containing 3, 000 pictures is created and manually annotated for network training. The experimental results show that the algorithm can effectively detect the wearing of masks, and has achieved good detection results in natural scene videos.

    Related Articles | Metrics
    Research on UWB and LiDAR Fusion Positioning Algorithm in Indoor Environment
    LI Zhongdao, LIU Yuansheng, CHANG Feixiang, ZHANG Jun, LU Ming
    Computer Engineering and Applications    2021, 57 (6): 260-266.   DOI: 10.3778/j.issn.1002-8331.2005-0435
    Abstract212)      PDF(pc) (1110KB)(775)       Save

    At present, the data fusion of the global navigation satellite system and the light detection and ranging is widely used in the positioning system of autonomous vehicles, but in the indoor environment, the loss of satellite signals leads to low positioning accuracy or even positioning failure. Therefore, a fusion positioning algorithm based on Ultra-Wideband(UWB) and Light Detection and Ranging(LiDAR) is proposed. The algorithm is based on particle filter to solve the location data of two sensors by complementary fusion. The real-time positioning data of UWB is used to improve the positioning speed of LiDAR by providing the range of starting particles. The weight of particles is updated by solving the geometric distance between the LiDAR positioning information and the particles, thereby compensating the positioning error of UWB in a non-line-of-sight environment. An indoor test scene is built, and the fusion positioning algorithm is verified on the smart car platform. The experimental results show that this method is superior to the single sensor positioning scheme of UWB or LiDAR, and the vehicle can still obtain good positioning accuracy and real-time performance when the UWB line of sight is blocked or LiDAR matching fails.

    Related Articles | Metrics
    Review of Task Scheduling Methods in Cloud Computing Environment
    TIAN Zhuojing, HUANG Zhenchun, ZHANG Yinong
    Computer Engineering and Applications    2021, 57 (2): 1-11.   DOI: 10.3778/j.issn.1002-8331.2006-0259
    Abstract385)      PDF(pc) (702KB)(762)       Save

    With the rapid growth of application computing demand, heterogeneous computing resources continue to increase, task scheduling has become an important research problem in the field of cloud computing. Task scheduling is responsible for matching user tasks to appropriate virtual computing resources. The quality of the algorithm will directly affect the response time, makespan, energy consumption, cost, resource utilization and a series of performance indexes that are closely related to the economic interests of users and cloud service providers. This paper summarizes and discusses the research progress of task scheduling algorithm based on the characteristics of different cloud environments, aiming at independent task and scientific workflow. Firstly, it reviews the existing task scheduling types, scheduling mechanisms and their advantages and disadvantages. Secondly, task scheduling characteristics under single cloud environment, and inter-cloud environment such as hybrid cloud, multi-cloud and federated cloud are summarized, and schedule methods, optimization objectives, pros and cons of some typical relevant literatures are described. On this basis, the research status of task scheduling under various environments is discussed. Then, the scheduling optimization methods used in various environments are further sorted out to clarify their scope of use. Finally, it summarizes the whole paper and points out that it is necessary to pay more attention to the research of task scheduling in computing data intensive applications under inter-cloud environment.

    Related Articles | Metrics
    Survey on Research Status of Blind-Guiding Robots
    WU Zhaohan, RONG Xuewen, FAN Yong
    Computer Engineering and Applications    2020, 56 (14): 1-13.   DOI: 10.3778/j.issn.1002-8331.2001-0298
    Abstract447)      PDF(pc) (1778KB)(754)       Save

    Blind people are the vulnerable groups in human society, and the number of them is increasing year by year. Providing safe, reliable, intelligent and efficient travel guarantees for the blind is an important sign of social progress. The research results of white-cane-based, wearable, handheld, intelligent-terminal-based and mobile blind-guiding robots are introduced. The research status of the common key technology of environment detection, positioning and navigation, and human-computer interaction of blind-guiding robots is summarized. Finally, the future development trend of blind-guiding robots is forecasted based on the latest technologies such as cloud platform and mobile communication.

    Related Articles | Metrics
    Improved YOLOv3 Object Detection Algorithm:Combining GIoU and Focal loss
    ZOU Chengming, XUE Ronggang
    Computer Engineering and Applications    2020, 56 (24): 214-222.   DOI: 10.3778/j.issn.1002-8331.1912-0428
    Abstract227)      PDF(pc) (1150KB)(753)       Save

    As an object detection algorithm, YOLOv3 can achieve fast detection speed and high detection accuracy. However, YOLOv3 has the problems of low detection accuracy for small objects and inaccurate boundary box location. In this paper, an improved YOLOv3 object detection algorithm is proposed, which increases the bypass connection between residual blocks in darknet-53 and further reuses the features to extract more feature information. At the same time, GIoU loss is used as the loss of bounding box, which can make the network optimize in the direction of high overlap between prediction box and ground truth. In addition, in order to reduce the error caused by the imbalance of positive and negative samples, Focal loss is added to the loss function. The experimental results on PASCAL VOC and COCO datasets show that the improved YOLOv3 algorithm can improve the accuracy of object detection on the premise of real-time performance. Compared with YOLOv3, the improved YOLOv3 algorithm reaches 83.7 mAP(IoU=0.5) in PASCAL VOC 2007 test and improves 2.27mAP(IoU [0.5, 0.95])on the COCO datasets.

    Related Articles | Metrics
    Survey of Multimodal Deep Learning
    SUN Yingying, JIA Zhentang, ZHU Haoyu
    Computer Engineering and Applications    2020, 56 (21): 1-10.   DOI: 10.3778/j.issn.1002-8331.2002-0342
    Abstract757)      PDF(pc) (769KB)(731)       Save

    Modal refers to the way people receive information, including hearing, vision, smell, touch and other ways. Multimodal learning refers to learning better feature representation by using the complementarity between multimodes and eliminating the redundancy between them. The purpose of multimodal learning is to build a model that can deal with and correlate information from multiple modes. It is a dynamic multidisciplinary field, with increasing importance and great potential. At present, the popular research direction is multimodal learning among image, video, audio and text. This paper focuses on the application of multimodality in audio-visual speech recognition, image and text emotion analysis, collaborative annotation and other practical levels, as well as the application in the core level of matching and classification, alignment representation learning, and gives an explanation for the core issues of multimodal learning:matching and classification, alignment representation learning. Finally, the common data sets in multimodal learning are introduced, and the development trend of multimodal learning in the future is prospected.

    Related Articles | Metrics
    Review of Research on Joint Intent Detection and Semantic Slot Filling in End to End Dialogue System
    WANG Kun, LIN Min, LI Yanling
    Computer Engineering and Applications    2020, 56 (14): 14-25.   DOI: 10.3778/j.issn.1002-8331.2004-0142
    Abstract211)      PDF(pc) (771KB)(700)       Save

    The end-to-end dialogue system based on deep learning has become hotspot research in academia and industry, because it has the advantages of strong generalization ability, few training parameters, and good performance. The results of intent detection and semantic slot filling are critical to the performance of the dialogue system. This paper introduces the mainstream methods of joint intent detection and semantic slot filling in an end-to-end task-oriented dialogue system. It not only summarizes the advantages of attention mechanism and Transformer model compared to recurrent neural network and long short-term memory network in capturing long-term dependencies, but also introduces the problem of imperfect capture of word position information caused by parallel processing. Then, it analyzes the improvement of capsule neural networks capturing small probability semantic information and keeping feature integrity compared to convolution neural networks. Furthermore, it mainly introduces the joint recognition method based on the BERT(Bidirectional Encoder Representations from Transformers) model, which can not only process in parallel but also solve the problem of polysemy, which is the best method at present. Finally it discusses and analyzes the future research direction.

    Related Articles | Metrics
    Product Traceability System of Automobile Supply Chain Based on Block Chain
    LI Baodong, YE Chunming
    Computer Engineering and Applications    2020, 56 (24): 35-42.   DOI: 10.3778/j.issn.1002-8331.2006-0390
    Abstract233)      PDF(pc) (928KB)(691)       Save

    Blockchain has the characteristics of openness, transparency, non-tampering and easy traceability as an emerging distributed database technology. It has a good fit with the traceability of supply chain products. Aiming at the problems of automobile supply chain, such as lack of trust, difficulty of traceability and low efficiency of information sharing, a product traceability system of automobile supply chain is designed based on block chain technology. Ethereum is selected as the development platform of the system, and functional modules such as authorization management, information input, traceability transfer and chain query are designed. Smart contracts are designed according to functional requirements, and processing links of sensitive data are added. The system is equipped with four participants, namely, raw material suppliers, parts suppliers, OEMs and distributors. The system provides the traceability interface between regulatory authorities and consumers. Finally, an example is given to demonstrate the traceability system. The research shows that the blockchain traceability system has obvious advantages over the traditional traceability method in terms of product data security and traceability efficiency.

    Related Articles | Metrics
    Research Progress of Multi-label Text Classification
    HAO Chao, QIU Hangping, SUN Yi, ZHANG Chaoran
    Computer Engineering and Applications    2021, 57 (10): 48-56.   DOI: 10.3778/j.issn.1002-8331.2101-0096
    Abstract722)      PDF(pc) (906KB)(688)       Save

    As a basic task in natural language processing, text classification has been studied in the 1950s. Now the single-label text classification algorithm has matured, but there is still a lot of improvement on multi-label text classification. Firstly, the basic concepts and basic processes of multi-label text classification are introduced, including data set acquisition, text preprocessing, model training and prediction results. Secondly, the methods of multi-label text classification are introduced. These methods are mainly divided into two categories:traditional machine learning methods and the methods based on deep learning. Traditional machine learning methods mainly include problem transformation methods and algorithm adaptation methods. The methods based on deep learning use various neural network models to handle multi-label text classification problems. According to the model structure, they are divided into multi-label text classification methods based on CNN structure, RNN structure and Transfomer structure. The data sets commonly used in multi-label text classification are summarized. Finally, the future development trend is summarized and analyzed.

    Related Articles | Metrics
    Review of Development and Application of Artificial Neural Network Models
    ZHANG Chi, GUO Yuan, LI Ming
    Computer Engineering and Applications    2021, 57 (11): 57-69.   DOI: 10.3778/j.issn.1002-8331.2102-0256
    Abstract537)      PDF(pc) (781KB)(665)       Save

    Artificial neural networks are increasingly closely related to other subject areas. People solve problems in various fields by exploring and improving the layer structure of artificial neural networks. Based on the analysis of artificial neural networks related literature, this paper summarizes the history of artificial neural network growth and presents relevant principles of artificial neural networks based on the development of neural networks, including multilayer perceptron, back-propagation algorithm, convolutional neural network and recurrent neural network, explains the classic convolutional neural network model in the development of the convolutional neural network and the widely used variant network structure in the recurrent neural network, reviews the application of each artificial neural network algorithm in related fields, summarizes the possible direction of development of the artificial neural network.

    Related Articles | Metrics
    Review of Pedestrian Detection with Occlusion
    CHEN Ning, LI Menglu, YUAN Hao, LI Yunhong, YANG Di, LIU Zhijian
    Computer Engineering and Applications    2020, 56 (16): 13-20.   DOI: 10.3778/j.issn.1002-8331.2003-0116
    Abstract285)      PDF(pc) (988KB)(662)       Save

    Pedestrian detection has been widely applied in intelligent traffic, intelligent monitoring, unmanned driving, pedestrian analysis and other fields. With the development of technology, the accuracy of pedestrian detection technology has become increasingly high. This paper summarizes the research progress of pedestrian detection technology under occlusion. Firstly, according to different occlusion, it can be divided into reasonable-occlusion caused by non-target and reasonable-crowd caused by target to be detected. Secondly, this paper summarizes the traditional method and deep learning method to deal with occlusion. The main ideas and core problems of each method model are analyzed and discussed. Finally, this paper gives an outlook on the problems to be solved in the future development of pedestrian detection technology under occlusion.

    Related Articles | Metrics
    Influence of Identity Loss on CycleGAN
    LIU Huachao, ZHANG Junran, LIU Yunfei
    Computer Engineering and Applications    2020, 56 (22): 217-223.   DOI: 10.3778/j.issn.1002-8331.1910-0405
    Abstract431)      PDF(pc) (1862KB)(648)       Save

    In the field of image generation, traditional image style transfer requires conversion between two paired images. CycleGAN(Cycle Generative Adversarial Network) has achieved good results by virtue of its ability to generate images for unpaired images, has become a research hotspot in the field of image generation. However, the classical CycleGAN generator cannot accurately identify the specific transformation domain and irrelevant domain of the image. Thus, it has the shortcoming of the arbitrary transformation of image irrelevant domain features, and the distortion of the generated image. For the above questions, identity loss is introduced to constrain the feature recognition of generator, L1 loss is used to guarantee that the converted image corresponds to the pixel level of the original image, effectively improve the problem, and make the generated image clearer. Meanwhile, by adjusting the ratio of identity loss [μ], the identity loss under different [μ] values is further analyzed, and the change of the loss of  CycleGAN and its effect on the quality of the generated image are discuessed. Finally, the selection strategy of proportional superparameter [μ] value of identity loss is given.

    Related Articles | Metrics
    Road Small Target Detection Algorithm Based on Improved YOLO V3
    YUE Xiaoxin, JIA Junxia, CHEN Xidong, LI Guang’an
    Computer Engineering and Applications    2020, 56 (21): 218-223.   DOI: 10.3778/j.issn.1002-8331.2005-0340
    Abstract190)      PDF(pc) (810KB)(641)       Save

    Aiming at the problems of poor performance and high missed detection rate when the general target detection algorithm detects small targets, a road small target detection algorithm based on improved YOLO V3 is proposed. The clustering algorithm in the original YOLO V3 algorithm network model is optimized, the DBSCAN+K-Means clustering algorithm is used to cluster the training dataset, more appropriate Anchor Boxs are selected to improve the detection average precision and speed. At the same time, the Focal Loss loss function is introduced to replace the loss function in the original network model to form an improved YOLO V3 algorithm. Compared with other target detection algorithms on the KITTI dataset for person detection, it is found that the improved YOLO V3 algorithm can effectively reduce the missed detection rate of small target detection, and greatly improve the average precision and detection speed. Experimental results show that, on the KITTI dataset, the average precision of the improved YOLO V3 algorithm reaches 92.43%, which is 2.36% higher than the unimproved YOLO V3 algorithm, and the detection speed reaches 44.52 frames per second.

    Related Articles | Metrics
    3D Point Cloud Classification Based on K-means Clustering
    MA Jinghui, PAN Wei, WANG Ru
    Computer Engineering and Applications    2020, 56 (17): 181-186.   DOI: 10.3778/j.issn.1002-8331.1909-0305
    Abstract242)      PDF(pc) (826KB)(618)       Save

    Aiming at the problem that the performance of 3D point-cloud classification algorithm is affected by point-cloud sparsity and disorder, this paper proposes an improved algorithm based on PointNet which is proposed in 2018. Firstly, during the point-cloud preprocessing, redundant data are removed from dense point-clouds to reduce the complexity of subsequent work. And at the same time, triangle interpolation is used in the sparse point-cloud data to make the classification more precise. Secondly, it uses K-means algorithm to cluster the preprossed data and put them through the PointNet network in parallel. The distribution characteristics of point-cloud can be obtained by this way. Experiments are made on ModelNet10/40 and compared with some popular classification algorithms based on deep learning. The results show that the performance of this new algorithm is the best in the above algorithms while the training time is greatly reduced.

    Related Articles | Metrics
    Review of Pre-training Models for Natural Language Processing
    YU Tongrui, JIN Ran, HAN Xiaozhen, LI Jiahui, YU Ting
    Computer Engineering and Applications    2020, 56 (23): 12-22.   DOI: 10.3778/j.issn.1002-8331.2006-0040
    Abstract602)      PDF(pc) (689KB)(597)       Save

    In recent years, deep learning technology has been advancing, pre-training technology for deep learning brings natural language processing into a new era. Pre-training model aims to how to make pre-trained model stay in good initial state and achieve better performances in subsequent downstream tasks. This paper firstly introduces pre-training technology and its development history. And then, this paper further classifies it into the following two types, namely probability-statistics-based traditional model and deep-learning-based new model, according to different features of pre-training models to conduct corresponding detailed introductions. Firstly, it briefly analyzes the characteristics and limitations of today’s pre-training models and highlights the existing deep-learning-based pre-training models. And based on their performances in downstream tasks, it gives necessary comparisons and assessments accordingly. Finally, it combs out a series of whole-new pre-training models with instructive significances,briefly describes corresponding feasible improvement mechanisms and the performance enhancements achieved in downstream tasks, summarizes the problems existing therein, as well as prospectes its development trend in near future.

    Related Articles | Metrics
    Survey of Object Detection Algorithms for Deep Convolutional Neural Networks
    HUANG Jian, ZHANG Gang
    Computer Engineering and Applications    2020, 56 (17): 12-23.   DOI: 10.3778/j.issn.1002-8331.2005-0021
    Abstract434)      PDF(pc) (1058KB)(594)       Save

    Object detection is one of the core tasks in computer vision, which is widely used in intelligent video monitoring, automatic monitoring, industrial detection and other fields. In recent years, with the rapid development of deep learning, the object detection algorithm based on deep convolutional neural network has gradually replaced the traditional object detection algorithm and become the mainstream algorithm in this field. Firstly, the common data sets and performance evaluation indexes of the object detection algorithm are introduced, and then the development of the convolutional neural network is introduced. Then, the two-stage object detection algorithm and the single-stage object detection algorithm are analyzed and compared. Finally, the future development of the object detection algorithm based on the deep convolutional neural network is prospected.

    Related Articles | Metrics
    Survey of Interpretability Research on Deep Learning Models
    ZENG Chunyan, YAN Kang, WANG Zhifeng, YU Yan, JI Chunmei
    Computer Engineering and Applications    2021, 57 (8): 1-9.   DOI: 10.3778/j.issn.1002-8331.2012-0357
    Abstract379)      PDF(pc) (677KB)(582)       Save

    With the characteristics of data-driven learning, deep learning technology has made great achievements in the fields of natural language processing, image processing, and speech recognition. However, due to the deep learning model featured by deep networks, many parameters, high complexity and other characteristics, the decisions and intermediate processes made by the model are difficult for humans to understand. Therefore, exploring the interpretability of deep learning has become a new topic in the current artificial intelligence field. This review takes the interpretability of deep learning models as the research object and summarizes its progress. Firstly, the main interpretability methods are summarized and analyzed from four aspects:self-explanatory model, model-specific explanation, model-agnostic explanation, and causal interpretability. At the same time, it enumerates the application of interpretability related technologies, and finally discusses the existing problems of current interpretability research to promote the further development of the deep learning interpretability research framework.

    Related Articles | Metrics
    Survey of Entity Relation Extraction
    WANG Chuandong, XU Jiao, ZHANG Yong
    Computer Engineering and Applications    2020, 56 (12): 25-36.   DOI: 10.3778/j.issn.1002-8331.2003-0189
    Abstract285)      PDF(pc) (917KB)(567)       Save

    As an important part of information extraction, entity relation extraction can perform semantic analysis on smaller grained information and provide basic data support for more tasks. The development of relation extraction can be divided into two methods based on traditional machine learning and deep learning. In recent years, the research based on traditional machine learning has mainly focused on the combination of statistic-based and rule-based. The framework of deep learning has achieved abundant research results by introducing distant supervision, few-shot learning, attention mechanism, reinforcement learning and multi-instance multilabel. The development of entity relation extraction is reviewed and each model is analyzed. Combing the latest developments in deep learning methods, the development direction and trend of entity relationship extraction are prospected.

    Related Articles | Metrics
    Comparative Study of Several New Swarm Intelligence Optimization Algorithms
    LI Yali, WANG Shuqin, CHEN Qianru, WANG Xiaogang
    Computer Engineering and Applications    2020, 56 (22): 1-12.   DOI: 10.3778/j.issn.1002-8331.2006-0291
    Abstract961)      PDF(pc) (972KB)(566)       Save

    With the development of computer technology, algorithm technology is constantly and alternately being updated. In recent years, swarm intelligence algorithm has become more and more popular and received extensive attention and research, and has made progress in such fields as machine learning, process control and engineering prediction. Swarm intelligence optimization algorithm is a biological heuristic method, which is widely used in solving optimization problems. The traditional swarm intelligence algorithm provides some new ideas for solving some practical problems, but it also exposes some shortcomings in some experiments. In recent years, many scholars have proposed many new types of intelligent optimization algorithms. This paper selects the more typical swarm intelligence algorithms at home and abroad in recent years, such as Bat Algorithm(BA), Grey Wolf Optimization Algorithm(GWO), Dragonfly Algorithm(DA), Whale Optimization Algorithm(WOA), Grasshopper Optimization Algorithm(GOA) and Sparrows Search Algorithm(SSA), and further compares the experimental performance of these algorithms and the development potential by 22 standard CEC test functions from the convergence speed and accuracy, stability and so on, and the refinement analysis is carried out to compare and analyze the relevant improvement methods. Finally, the characteristics of swarm intelligence optimization algorithm are summarized and its development potential is discussed.

    Related Articles | Metrics
    Overview of Research on 3D Human Pose Estimation
    WANG Faming, LI Jianwei, CHEN Sixi
    Computer Engineering and Applications    2021, 57 (10): 26-38.   DOI: 10.3778/j.issn.1002-8331.2102-0039
    Abstract368)      PDF(pc) (1035KB)(560)       Save

    The 3D human pose estimation is essentially a classification and regression problem. It mainly estimates the 3D human pose from images. The 3D human pose estimation based on traditional methods and deep learning methods is the mainstream research method in this field. This paper follows the traditional methods to the deep learning methods to systematically introduce the 3D human posture estimation methods in recent years, and basically understands the traditional methods to obtain the many elements of the human posture through the generation and discrimination methods to complete the 3D human posture estimation. The 3D human pose estimation method based on deep learning mainly regresses the human pose information from the image features by constructing a neural network. It can be roughly divided into three categories:based on direct regression methods, based on 2D information methods, and based on hybrid methods. In the end, it summarizes the current research difficulties and challenges, and discusses the research trends.

    Related Articles | Metrics
    End to End Speech Recognition Based on ResNet-BLSTM
    HU Zhangfang, XU Xuan, FU Yaqin, XIA Zhiguang, MA Sudong
    Computer Engineering and Applications    2020, 56 (18): 124-130.   DOI: 10.3778/j.issn.1002-8331.1907-0019
    Abstract225)      PDF(pc) (880KB)(553)       Save

    In the end-to-end speech recognition model based on deep learning, the input of the model adopts fixed length speech frames, which results in the loss of time-domain information and part of high-frequency information, resulting in low recognition rate and at weak robust of system. According to the above problem, this paper proposes a model based on the ResNet and the BLSTM, the model uses the spectrogram as input, and simultaneously designs the parallel convolution layer in the residual network, extracts features of different scales, and then performs features fusion, and finally uses the connection timing classification method to classify and realize an end-to-end speech recognition model. The experimental results show that compared with the traditional end-to-end model, the WER of the model in this paper decreases by 2.52% on the Aishell-1 speech set, and the robustness is better.

    Related Articles | Metrics
    Survey on Question Classification Method in Question Answering System
    HAN Dongfang, Turdy Toheti, Askar Hamdulla
    Computer Engineering and Applications    2021, 57 (6): 10-21.   DOI: 10.3778/j.issn.1002-8331.2009-0211
    Abstract216)      PDF(pc) (729KB)(552)       Save

    As a high-level form of information retrieval, the Question Answering system(QA)can quickly and accurately provide users with the required information services. After giving a question, an accurate answer will be given accordingly, which makes it become a more and more attention research direction  in the field of natural language processing. Question Classification(QC)is the most important part of question analysis and processing in the QA, and its classification accuracy will directly affect the overall performance of the QA. In recent years, the rapid development of machine learning and deep learning technologies has greatly promoted the research and development of QC, which has strong feasibility and superiority in question classification. This paper summarizes and analyzes the domestic and foreign research status of QC, question classification standard system, question feature extraction, traditional machine learning classification methods and recently popular deep learning classification methods, and elaborates the current status of QC. This paper expounds research difficulties in QC, and makes preliminary prospects for future research and development directions.

    Related Articles | Metrics
    Research Status and Future Analysis of Capsule Neural Network
    HE Wenliang, ZHU Minling
    Computer Engineering and Applications    2021, 57 (3): 33-43.   DOI: 10.3778/j.issn.1002-8331.2009-0209
    Abstract697)      PDF(pc) (890KB)(549)       Save

    Nowadays, the rapid development of artificial intelligence technology has promoted the great progress of society. As an important part of artificial intelligence, deep learning has a very broad application prospect. In recent years, more and more experts and scholars have begun to study the related technologies in the field of deep learning. Two typical directions are natural language processing and computer vision. Among them, the development of computer vision strongly leads the progress in the field of deep learning. The application of convolutional neural network and a new neural network model in deep learning?capsule network and its dynamic routing algorithm are introduced, and their advantages and disadvantages are compared. Then, the application of capsule network is reviewed, and the application fields and advantages of capsule network are described in terms of image and text. Finally, the paper summarizes and looks forward to the possible improvement direction of capsule network.

    Related Articles | Metrics
    License Plate Location Detection Algorithm Based on Improved YOLOv3 in Complex Scenes
    MA Qiaomei, WANG Mingjun, LIANG Haoran
    Computer Engineering and Applications    2021, 57 (7): 198-208.   DOI: 10.3778/j.issn.1002-8331.2008-0137
    Abstract185)      PDF(pc) (1709KB)(549)       Save

    Aiming at the problem of the difficulty of license plate positioning, slow detection speed and low detection accuracy in complex scenes such as lighting, multi-vehicle and low resolution, an improved method based on YOLOv3 is proposed. Firstly, the label information of the example is clustered by K-means++ method to obtain a new anchor size. And then, the improved thin feature extraction network(DarkNet41) is used to improve the detection efficiency of the model and reduce computational consumption. Moreover, multi-scale feature fusion is improved from 3-scale prediction to 4-scale prediction and improved Inception-SE structure is added to the detection network to improve the accuracy of detection. Finally, CIoU is selected as a loss function. The data is enhanced with the Multi-Scale Retinex(MSR) algorithm. Experimental analysis shows that the improved algorithm’s mAP reaches 98.84% and the detection speed reaches 36.4 frame/s, which has better accuracy and real-time performance compared with the YOLOv3 model and other algorithms.

    Related Articles | Metrics
    Overview of Motion Deblurring Techniques for Images
    HUANG Zhengyuan, XIE Weicheng, HUANG Huaru, CAO Qian
    Computer Engineering and Applications    2020, 56 (24): 28-34.   DOI: 10.3778/j.issn.1002-8331.2006-0446
    Abstract249)      PDF(pc) (809KB)(542)       Save

    As the carrier of human information communication, image contains a large number of information elements. During the acquisition process, motion blur will occur due to camera shake, object displacement and other reasons, resulting in the image cannot convey information correctly. This kind of degraded image can be repaired and restored by image motion blur restoration technology, which is a hot spot in the field of computer vision and image processing. The image motion blur reduction method is divided into two types by whether the blurring kernel is known. This paper elaborates the concept of image motion blur reduction, summarizes the methods and research status of the image motion blur reduction technology in recent years, advantages and disadvantages of each method are summarized and the experimental results are compared to classical method, key problems of the fuzzy kernel estimation and fuzzy data set are analyzed and the direction of the development suggestions are given, the image motion blur reduction technology development trend is prospected.

    Related Articles | Metrics
    Research on Residual Networks for Image Classification
    ZHAO Liping, YUAN Xiao, ZHU Cheng, ZHAO Xiaoqi, YANG Shihu, LIANG Ping, LU Xiaoya, TAN Ying
    Computer Engineering and Applications    2020, 56 (20): 9-19.   DOI: 10.3778/j.issn.1002-8331.2005-0219
    Abstract304)      PDF(pc) (826KB)(540)       Save

    In recent years, with the continuous expansion of data sets and the continuous improvement of computer performance, the traditional image classification methods always lead to low accuracy. Because of their high accuracy and great convergence, the residual networks have become a key technical in the field of image classification, they are worthy of studying in detail. Each variant improves network performance by improving classification accuracy, reducing model complexity or reducing calculation amount. Firstly, This paper analyzes the advantages and disadvantages of each variant and the suggestions are given for the application of them. Then, the performance of each variant is compared intuitively from the three aspects of accuracy rate, parameter amount and calculation amount. Finally, the challenges and future development of residual networks are put forward.

    Related Articles | Metrics
    Review of Coordinated Control Algorithms for Dual-Arm Robots
    WANG Qi, MIN Huasong
    Computer Engineering and Applications    2021, 57 (1): 1-16.   DOI: 10.3778/j.issn.1002-8331.2008-0025
    Abstract373)      PDF(pc) (855KB)(536)       Save

    The dual-arm robot system is a research hotspot in the field of robot, especially with the limitation of single-arm robot in operation ability and control, the research focuses on the redundant dual-arm robot with coordinated operation ability. Firstly, the dual-arm operation is classified, and then it is analyzed from five aspects: the dual-arm coordinated motion mode, the dual-arm coordinated control problem, the sensing sensor, the imitation learning, and the human-computer interaction. Review from the current situation of kinematics, dynamics, it analyzes the differences and development of dual-arm coordinated control and single-arm control mode in constraint relationship, motion planning, coordinated control mode, and the application of sensing sensors and imitation learning in dual-arm coordinated control. Finally, the interaction mode of human-computer cooperation is analyzed, and the future research direction of dual-arm robot is prospected.

    Related Articles | Metrics