Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627500
Li Guanbin
An intelligent music playback control system, which can select the appropriate music to play according to the facial expression recognition result, is designed. The hardware system consists of PYNQ_Z2 development board, camera, HDMI display and audio. The software system performs face detection on the core chip of ARM + FPGA through Haar feature and Adaboost algorithm, and realizes expression recognition through LDA and K nearest neighbor algorithm. The expression recognition results are sub-classified by confidence, and the PID control algorithm is used to adjust the confidence to improve the robustness of the system during uneven illumination. The system has low delay, strong robustness and rich human-computer interaction, and has a wide range of application scenarios.
{"title":"A Music Playback Control System Based on Facial Expression Recognition","authors":"Li Guanbin","doi":"10.1109/icisfall51598.2021.9627500","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627500","url":null,"abstract":"An intelligent music playback control system, which can select the appropriate music to play according to the facial expression recognition result, is designed. The hardware system consists of PYNQ_Z2 development board, camera, HDMI display and audio. The software system performs face detection on the core chip of ARM + FPGA through Haar feature and Adaboost algorithm, and realizes expression recognition through LDA and K nearest neighbor algorithm. The expression recognition results are sub-classified by confidence, and the PID control algorithm is used to adjust the confidence to improve the robustness of the system during uneven illumination. The system has low delay, strong robustness and rich human-computer interaction, and has a wide range of application scenarios.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"215 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114008414","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627437
Yanyan Wang, Chunfang Li, Jiangnan Sun, Min Li, Jintian Yang
Driven by the application needs of computer-aided geometric design, computer animation, reverse engineering, medical diagnosis and entertainment industry, 3D point cloud data processing technology has attracted more and more attention. This article described some technology to process and manage the 3D point cloud data of athlete's feet, and the process of building a 3D data integrated system of athlete's foot information which consists of data acquisition, processing, upload, storage, visualization and authority control. Python Open3D is used to merge the foot mold point cloud data with base information, extract the foot mold point cloud data without base, denoise and generate mesh models. Three.JS technology is used to realize the smooth display and interaction of the model in Vue3 framework and Node.JS. The system can be used for assisting researchers in the foot modeling and sports biomechanical analysis of Winter Olympic athletes and sports students.
{"title":"An Athlete's Foot Data Platform with 3D Point Cloud Processing and Management Technology","authors":"Yanyan Wang, Chunfang Li, Jiangnan Sun, Min Li, Jintian Yang","doi":"10.1109/icisfall51598.2021.9627437","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627437","url":null,"abstract":"Driven by the application needs of computer-aided geometric design, computer animation, reverse engineering, medical diagnosis and entertainment industry, 3D point cloud data processing technology has attracted more and more attention. This article described some technology to process and manage the 3D point cloud data of athlete's feet, and the process of building a 3D data integrated system of athlete's foot information which consists of data acquisition, processing, upload, storage, visualization and authority control. Python Open3D is used to merge the foot mold point cloud data with base information, extract the foot mold point cloud data without base, denoise and generate mesh models. Three.JS technology is used to realize the smooth display and interaction of the model in Vue3 framework and Node.JS. The system can be used for assisting researchers in the foot modeling and sports biomechanical analysis of Winter Olympic athletes and sports students.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114582752","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627442
Zhebin Feng, Chunhua Wang, Wenqian Shang, Weiguo Lin
In traditional convolutional neural networks, the calculation process of input information is generally regarded as the process of feature extraction and representation. The effects of the models are closely related to the number of extracted features. In the research of this paper, the feature extraction process of neural network is regarded as a signal processing process. By using the feedback compensation mechanism of weak signal detection in the signal system, the output at the current time is fed back to the current input for information compensation, so as to achieve the effect of feature enhancement. This method is tested on MINIST data set and the experimental results show that the neural network with feedback compensation, without adding more parameters, can effectively improve the convergence speed of the model, reduce the fluctuation of loss function, and improve the accuracy. The comparison results show that the neural network with feedback compensation mechanism achieves the effect of feature enhancement.
{"title":"A Neural Network Feature Enhancement Method Based on Feedback Compensation Mechanism","authors":"Zhebin Feng, Chunhua Wang, Wenqian Shang, Weiguo Lin","doi":"10.1109/icisfall51598.2021.9627442","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627442","url":null,"abstract":"In traditional convolutional neural networks, the calculation process of input information is generally regarded as the process of feature extraction and representation. The effects of the models are closely related to the number of extracted features. In the research of this paper, the feature extraction process of neural network is regarded as a signal processing process. By using the feedback compensation mechanism of weak signal detection in the signal system, the output at the current time is fed back to the current input for information compensation, so as to achieve the effect of feature enhancement. This method is tested on MINIST data set and the experimental results show that the neural network with feedback compensation, without adding more parameters, can effectively improve the convergence speed of the model, reduce the fluctuation of loss function, and improve the accuracy. The comparison results show that the neural network with feedback compensation mechanism achieves the effect of feature enhancement.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"33 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115070421","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627410
Thami Batyashe, T. Iyamu
Evidently, information technology (IT) is increasingly a vital and pervasive instrument for any organisation's existence, competitiveness and sustainability. Despite this sheer importance, the deployment and use of IT solutions bring its own complexities, which have led to many studies. however, some of the challenges are routinises because the factors that influence them are not empirically known. A qualitative study was conducted, two cases where studied, from the private and public sectors. Semi-structured interviews were employed as a guide to collect data from the two cases. The data was interpretively analysed using a socio-technical theory, the five stages of innovation-decision process from the perspectives of diffusion of innovations. An operational architecture is proposed to assist business and IT practitioners to operationalise IT strategies in institutions.
{"title":"Operational Architecture Framework for Information Technology Solutions: Diffusion of Innovation Perspective","authors":"Thami Batyashe, T. Iyamu","doi":"10.1109/icisfall51598.2021.9627410","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627410","url":null,"abstract":"Evidently, information technology (IT) is increasingly a vital and pervasive instrument for any organisation's existence, competitiveness and sustainability. Despite this sheer importance, the deployment and use of IT solutions bring its own complexities, which have led to many studies. however, some of the challenges are routinises because the factors that influence them are not empirically known. A qualitative study was conducted, two cases where studied, from the private and public sectors. Semi-structured interviews were employed as a guide to collect data from the two cases. The data was interpretively analysed using a socio-technical theory, the five stages of innovation-decision process from the perspectives of diffusion of innovations. An operational architecture is proposed to assist business and IT practitioners to operationalise IT strategies in institutions.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"55 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"122800266","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627367
Guozhen Wang, Wei-hua Niu, Jianping Zhao
Aiming at the problem that the periodic impact component extraction is difficult under the random impact interference in the actual vibration signal, a method of extracting the periodic impact component of the signal based on improved matching pursuit is proposed. The method proposed in this paper constructs the periodic impact atomic library, and the segmentation inner product method based on statistical improvement is used. The proposed method can effectively extract the periodic impact component of rotating machinery under random impact, and eliminate the impact of random impact. The simulation results show that the proposed method can effectively extract the periodic impact components in the signal under random impact interference, which proves the effectiveness and engineering practicability of the method.
{"title":"A Signal Periodic Impact Component Extraction Method Based on Matching Pursuit","authors":"Guozhen Wang, Wei-hua Niu, Jianping Zhao","doi":"10.1109/icisfall51598.2021.9627367","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627367","url":null,"abstract":"Aiming at the problem that the periodic impact component extraction is difficult under the random impact interference in the actual vibration signal, a method of extracting the periodic impact component of the signal based on improved matching pursuit is proposed. The method proposed in this paper constructs the periodic impact atomic library, and the segmentation inner product method based on statistical improvement is used. The proposed method can effectively extract the periodic impact component of rotating machinery under random impact, and eliminate the impact of random impact. The simulation results show that the proposed method can effectively extract the periodic impact components in the signal under random impact interference, which proves the effectiveness and engineering practicability of the method.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"120 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124787269","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627362
L. Bai, Pengcheng Wen, Yulin Hai, Ze Gao, Taoran Cheng, Heng Wang
This paper conducts research on the applicability technology of intelligent processors in embedded devices. From the aspects of the complexity of intelligent tasks, the real-time performance and accuracy requirements of intelligent tasks, the high-performance density requirements of the embedded system and the working environment requirements of embedded devices, the relevant characteristics of intelligent applications in the embedded environment are analyzed. Based on the above analysis, a series of testing indexes for the applicability of intelligent processors for embedded environments are proposed, including support for different types of intelligent algorithms, processing performance, processing accuracy, power consumption, and working environment. Using typical deep neural network models, the applicability of a certain type of domestic intelligent processor is tested and analyzed to verify the validity of the proposed indexes.
{"title":"Applicability Testing Technique of Intelligent Processor for Embedded Computing System","authors":"L. Bai, Pengcheng Wen, Yulin Hai, Ze Gao, Taoran Cheng, Heng Wang","doi":"10.1109/icisfall51598.2021.9627362","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627362","url":null,"abstract":"This paper conducts research on the applicability technology of intelligent processors in embedded devices. From the aspects of the complexity of intelligent tasks, the real-time performance and accuracy requirements of intelligent tasks, the high-performance density requirements of the embedded system and the working environment requirements of embedded devices, the relevant characteristics of intelligent applications in the embedded environment are analyzed. Based on the above analysis, a series of testing indexes for the applicability of intelligent processors for embedded environments are proposed, including support for different types of intelligent algorithms, processing performance, processing accuracy, power consumption, and working environment. Using typical deep neural network models, the applicability of a certain type of domestic intelligent processor is tested and analyzed to verify the validity of the proposed indexes.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"25 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125718127","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/ICISFall51598.2021.9627473
Xue Jiang, Yuan Zhang
This paper presents a perceptual content-aware bitrate adaptation algorithm for the HTTP streaming services. Compared with the traditional throughput-based and buffer-based algorithms, the impact of visual perception on user's Quality of Experience (QoE) has also been considered. We model the content-aware bitrate adaptation problem into a Markov Decision Process (MDP) and develop a segmented value iteration method to solve this problem. We have integrated this adaptive algorithm into dash.js, on which we can compare our approach with the default throughput-based algorithm and well-known BOLA algorithm. The results have shown that our algorithm can not only reach the higher average quality on the premise of maintaining fluency but also enable the scenes with higher attention to obtain higher quality, thus ultimately improve QoE.
{"title":"Perceptual Content-Aware Bitrate Adaptation for HTTP Streaming using Markov Decision Process","authors":"Xue Jiang, Yuan Zhang","doi":"10.1109/ICISFall51598.2021.9627473","DOIUrl":"https://doi.org/10.1109/ICISFall51598.2021.9627473","url":null,"abstract":"This paper presents a perceptual content-aware bitrate adaptation algorithm for the HTTP streaming services. Compared with the traditional throughput-based and buffer-based algorithms, the impact of visual perception on user's Quality of Experience (QoE) has also been considered. We model the content-aware bitrate adaptation problem into a Markov Decision Process (MDP) and develop a segmented value iteration method to solve this problem. We have integrated this adaptive algorithm into dash.js, on which we can compare our approach with the default throughput-based algorithm and well-known BOLA algorithm. The results have shown that our algorithm can not only reach the higher average quality on the premise of maintaining fluency but also enable the scenes with higher attention to obtain higher quality, thus ultimately improve QoE.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126726119","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627365
Joseph R. Fanfarelli
Computational thinking is an important skill for solving complex problems, including processes such as decomposition, pattern recognition, abstraction, and algorithmic design. Game-based learning has recently seen an increase in prevalence for teaching computational thinking, making games an important topic of study. However, there is currently no validated tool for assessing Computational Thinking (CT) that performs reliably across disciplines and age groups. In the absence of such a tool, this paper examines several software testing methods for the evaluation of CT pedagogy effectiveness within serious games. Namely, it makes recommendations for the application of standardized questionnaires, think-aloud testing, and automated data logging for evaluating games that promote CT learning. It concludes with a potential use case to demonstrate how the methods can be combined to achieve a granular and actionable understanding of a complex CT assessment problem and its causes.
{"title":"Assessing Computational Thinking Pedagogy in Serious Games Through Questionnaires, Think-aloud Testing, and Automated Data Logging","authors":"Joseph R. Fanfarelli","doi":"10.1109/icisfall51598.2021.9627365","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627365","url":null,"abstract":"Computational thinking is an important skill for solving complex problems, including processes such as decomposition, pattern recognition, abstraction, and algorithmic design. Game-based learning has recently seen an increase in prevalence for teaching computational thinking, making games an important topic of study. However, there is currently no validated tool for assessing Computational Thinking (CT) that performs reliably across disciplines and age groups. In the absence of such a tool, this paper examines several software testing methods for the evaluation of CT pedagogy effectiveness within serious games. Namely, it makes recommendations for the application of standardized questionnaires, think-aloud testing, and automated data logging for evaluating games that promote CT learning. It concludes with a potential use case to demonstrate how the methods can be combined to achieve a granular and actionable understanding of a complex CT assessment problem and its causes.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"928 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127017766","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627458
Zhigang Song, Kang Song, Nanchang Cheng, Jiao Li, Wenqian Shang, Yuanjun Zou
This paper mainly studies the dynamic identification of hot topics and their trend prediction: the identification methods of hot topics are studied from the two dimensions of content and form; the trend prediction method is completed from the two dimensions of media attention and emotional tendency. This paper develops a hot topic recognition method based on formal feature ranking. This paper compares the advantages and disadvantages of traditional methods, and proposes a high-frequency co-occurrence clustering strategy based on minimum similarity, which effectively solves the timeliness requirements of real-time dynamic hot spot recognition. Based on the recognition of hot topics, this article will jointly complete the trend prediction of hot topics from the changes in media attention and emotional orientation. We have encapsulated the hot topic recognition method based on high-frequency co-occurrence into a module and applied it in the national language and writing public opinion monitoring system.
{"title":"Research on Internet Public Opinion Recognition Method Based on High Frequency Co-occurrence","authors":"Zhigang Song, Kang Song, Nanchang Cheng, Jiao Li, Wenqian Shang, Yuanjun Zou","doi":"10.1109/icisfall51598.2021.9627458","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627458","url":null,"abstract":"This paper mainly studies the dynamic identification of hot topics and their trend prediction: the identification methods of hot topics are studied from the two dimensions of content and form; the trend prediction method is completed from the two dimensions of media attention and emotional tendency. This paper develops a hot topic recognition method based on formal feature ranking. This paper compares the advantages and disadvantages of traditional methods, and proposes a high-frequency co-occurrence clustering strategy based on minimum similarity, which effectively solves the timeliness requirements of real-time dynamic hot spot recognition. Based on the recognition of hot topics, this article will jointly complete the trend prediction of hot topics from the changes in media attention and emotional orientation. We have encapsulated the hot topic recognition method based on high-frequency co-occurrence into a module and applied it in the national language and writing public opinion monitoring system.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"2012 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127376755","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-10-13DOI: 10.1109/icisfall51598.2021.9627375
Richard Xue, Longquan Jiang, Peng Wang, Rui Feng, Fei Shan
Currently, manual analysis performed by professional radiologists is required for COVID-19 diagnosis given the patient's chest Computed Tomography (CT) images, but this process is inefficient and costly. Deep learning methods can provide computer vision-based solutions to help guide radiologists perform faster and more accurate diagnosis. However, current well performed methods require training on large and balanced datasets with pixel level lung lesion annotations, both of which are not easily accessible. Moreover, visual similarities between COVID-19 and other pneumonia in CT scans make it difficult to learn their distinguishing features. To address these issues, we propose a novel weakly-supervised deep learning model, named Multi-DeepNet, that can be well trained to perform fine-grained classification on small and imbalanced datasets. Specifically, a multi-task pre-training module is introduced to better extract distinguishing features between COVID-19 and other similar pneumonia. Furthermore, a multi-view-oriented classifier is proposed to extract complimentary information from the axial, coronal and sagittal planes. Experimental results demonstrate that our Multi-DeepNet achieves superior sensitivities, specificity, and accuracies compared to state-of-the-art methods.
{"title":"Multi-DeepNet: A Novel Weakly-Supervised Multi-Task and Multi-View-Oriented Convolution Neural Network for COVID-19 Diagnosis from CT Images","authors":"Richard Xue, Longquan Jiang, Peng Wang, Rui Feng, Fei Shan","doi":"10.1109/icisfall51598.2021.9627375","DOIUrl":"https://doi.org/10.1109/icisfall51598.2021.9627375","url":null,"abstract":"Currently, manual analysis performed by professional radiologists is required for COVID-19 diagnosis given the patient's chest Computed Tomography (CT) images, but this process is inefficient and costly. Deep learning methods can provide computer vision-based solutions to help guide radiologists perform faster and more accurate diagnosis. However, current well performed methods require training on large and balanced datasets with pixel level lung lesion annotations, both of which are not easily accessible. Moreover, visual similarities between COVID-19 and other pneumonia in CT scans make it difficult to learn their distinguishing features. To address these issues, we propose a novel weakly-supervised deep learning model, named Multi-DeepNet, that can be well trained to perform fine-grained classification on small and imbalanced datasets. Specifically, a multi-task pre-training module is introduced to better extract distinguishing features between COVID-19 and other similar pneumonia. Furthermore, a multi-view-oriented classifier is proposed to extract complimentary information from the axial, coronal and sagittal planes. Experimental results demonstrate that our Multi-DeepNet achieves superior sensitivities, specificity, and accuracies compared to state-of-the-art methods.","PeriodicalId":240142,"journal":{"name":"2021 IEEE/ACIS 20th International Fall Conference on Computer and Information Science (ICIS Fall)","volume":"20 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-10-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134641547","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：