Pub Date : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941310
Jin Matsuzaki, K. Sakakibara, Masaki Nakamura
This paper addresses the problem of scheduling machining operations in a highly automated manufacturing environment, taking into account the work styles of workers. In actual manufacturing, many issues must be taken into accounts, such as constraints related to the works to be machined in the machining schedule and the conditions of workers. To derive good solutions to such a large-scale problem with many constraints in a realistic amount of computing time, we develop an optimization technique based on the MIP-based large neighborhood local search method for the machining scheduling problem. Then, computer experiments are conducted on a problem created concerning actual machining requirements to verify the validity of the proposed method.
{"title":"The MIP-Based Large Neighborhood Local Search Method for Large-Scale Optimization Problems with Many Constraints: Application to the Machining Scheduling","authors":"Jin Matsuzaki, K. Sakakibara, Masaki Nakamura","doi":"10.1109/ICMLC56445.2022.9941310","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941310","url":null,"abstract":"This paper addresses the problem of scheduling machining operations in a highly automated manufacturing environment, taking into account the work styles of workers. In actual manufacturing, many issues must be taken into accounts, such as constraints related to the works to be machined in the machining schedule and the conditions of workers. To derive good solutions to such a large-scale problem with many constraints in a realistic amount of computing time, we develop an optimization technique based on the MIP-based large neighborhood local search method for the machining scheduling problem. Then, computer experiments are conducted on a problem created concerning actual machining requirements to verify the validity of the proposed method.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"88 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127116594","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941331
T. Ueyama, Yohei Kumabe, K. Oe, T. Fukui, T. Niikura, R. Kuroda, Masakazu Morimoto, N. Yagi, Y. Hata
In this paper, we simulate the fracture healing process using ultrasound and Bone Mineral Density (BMD). The frequency component of the reflected wave from the rat's bone is used. A hole was drilled in the center of the rat's femur to simulate a fracture. Firstly, the frequency response is obtained by adapting a Fast Fourier Transform to the resulting reflected wave, which is then cross spectrum to highlight characteristic frequencies. Next, we use the frequency and BMD healthy bone data to construct a pseudo-individual without considering overlap. Finally, we determine the degree of healing process for each individual. In our previous studies, it has the lack of reliability since there was only one data set that was set to be a bone hole was a problem, so the objective was to increase the number of data and improve reliability. The reliability of bone hole selection is demonstrated by comparing data increased frequencies data for pseudo-individuals with increased data frequencies to the healing process of pseudo-individuals used BMD from previous studies.
{"title":"Simulation Of Bone Fracture Healing Process Using Ultrasound And BMD Data","authors":"T. Ueyama, Yohei Kumabe, K. Oe, T. Fukui, T. Niikura, R. Kuroda, Masakazu Morimoto, N. Yagi, Y. Hata","doi":"10.1109/ICMLC56445.2022.9941331","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941331","url":null,"abstract":"In this paper, we simulate the fracture healing process using ultrasound and Bone Mineral Density (BMD). The frequency component of the reflected wave from the rat's bone is used. A hole was drilled in the center of the rat's femur to simulate a fracture. Firstly, the frequency response is obtained by adapting a Fast Fourier Transform to the resulting reflected wave, which is then cross spectrum to highlight characteristic frequencies. Next, we use the frequency and BMD healthy bone data to construct a pseudo-individual without considering overlap. Finally, we determine the degree of healing process for each individual. In our previous studies, it has the lack of reliability since there was only one data set that was set to be a bone hole was a problem, so the objective was to increase the number of data and improve reliability. The reliability of bone hole selection is demonstrated by comparing data increased frequencies data for pseudo-individuals with increased data frequencies to the healing process of pseudo-individuals used BMD from previous studies.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"5 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130302481","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941335
Xin Cheng, Meiqi Wang, Yuanyuan Shi, Jun Lin, Zhongfeng Wang
Model compression is one of the most preferred techniques for efficiently deploying deep neural networks (DNNs) on resource- constrained Internet of Things (IoT) platforms. However, the simply compressed model is often vulnerable to adversarial attacks, leading to a conflict between robustness and efficiency, especially for IoT devices exposed to complex real-world scenarios. We, for the first time, address this problem by developing a novel framework dubbed Magical-Decomposition to simultaneously enhance both robustness and efficiency for hardware. By leveraging a hardware-friendly model compression method called singular value decomposition, the defending algorithm can be supported by most of the existing DNN hardware accelerators. To step further, by using a recently developed DNN interpretation tool, the underlying scheme of how the adversarial accuracy can be increased in the compressed model is highlighted clearly. Ablation studies and extensive experiments under various attacks/models/datasets consistently validate the effectiveness and scalability of the proposed framework.
{"title":"Magical-Decomposition: Winning Both Adversarial Robustness and Efficiency on Hardware","authors":"Xin Cheng, Meiqi Wang, Yuanyuan Shi, Jun Lin, Zhongfeng Wang","doi":"10.1109/ICMLC56445.2022.9941335","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941335","url":null,"abstract":"Model compression is one of the most preferred techniques for efficiently deploying deep neural networks (DNNs) on resource- constrained Internet of Things (IoT) platforms. However, the simply compressed model is often vulnerable to adversarial attacks, leading to a conflict between robustness and efficiency, especially for IoT devices exposed to complex real-world scenarios. We, for the first time, address this problem by developing a novel framework dubbed Magical-Decomposition to simultaneously enhance both robustness and efficiency for hardware. By leveraging a hardware-friendly model compression method called singular value decomposition, the defending algorithm can be supported by most of the existing DNN hardware accelerators. To step further, by using a recently developed DNN interpretation tool, the underlying scheme of how the adversarial accuracy can be increased in the compressed model is highlighted clearly. Ablation studies and extensive experiments under various attacks/models/datasets consistently validate the effectiveness and scalability of the proposed framework.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"14 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116176054","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941293
Xiaojian Weng, Xudong Lin, S. Zhao
Price movements in the stock market affect all aspects of the social economy, and forecasting stock prices is of great importance. Traditional stock forecasting models are based on statistical regression models, which are difficult to characterize the influential relationships between multiple variables and predict stock price trends with large errors. In recent years, with the development of neural networks, neural networks have become a common method for stock forecasting, which include Back Propagation (BP) neural network, Convolutional Neural Networks (CNN), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) neural network. However, most of the previous stock price prediction models only use the basic stock market data, ignoring the influence of stock market investor sentiment on stock prices. A new stock price prediction model is proposed to address the above problems. First, the investor sentiment before the stock opening is calculated by fine-tuning the BERT model, then the calculated investor sentiment and the basic stock quotation data are aggregated, and finally the LSTM model is used to predict the closing price of the next stock trading day. We validate the effectiveness of the model on a real dataset of three Chinese listed companies.
{"title":"Stock Price Prediction Based On Lstm And Bert","authors":"Xiaojian Weng, Xudong Lin, S. Zhao","doi":"10.1109/ICMLC56445.2022.9941293","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941293","url":null,"abstract":"Price movements in the stock market affect all aspects of the social economy, and forecasting stock prices is of great importance. Traditional stock forecasting models are based on statistical regression models, which are difficult to characterize the influential relationships between multiple variables and predict stock price trends with large errors. In recent years, with the development of neural networks, neural networks have become a common method for stock forecasting, which include Back Propagation (BP) neural network, Convolutional Neural Networks (CNN), Recurrent Neural Network (RNN), and Long Short-Term Memory (LSTM) neural network. However, most of the previous stock price prediction models only use the basic stock market data, ignoring the influence of stock market investor sentiment on stock prices. A new stock price prediction model is proposed to address the above problems. First, the investor sentiment before the stock opening is calculated by fine-tuning the BERT model, then the calculated investor sentiment and the basic stock quotation data are aggregated, and finally the LSTM model is used to predict the closing price of the next stock trading day. We validate the effectiveness of the model on a real dataset of three Chinese listed companies.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"128 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126266430","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941332
Sumindar Kaur Saini, Mankaran Singh Ghumman
The traffic menace in India’s metropolitan cities causes many travelers to suffer daily. In traffic control, simple and old forms of signal controllers, known as electro-mechanical signal controllers, are used till-date which use dial timers that have fixed, signalized intersection time plans. As the time is fixed, the people in the lane with the greatest number of vehicles must wait the most, leading to wastage of time, money, and natural resources such as petrol and diesel. The proposed system is a traffic light system with feedback in real-time. The vehicles present in a specific lane are detected using a camera and then the deep learning algorithm, YOLO (You Only Look Once) detects the total number of vehicles in a lane which is used for feedback control of the lights. The traffic lights controller changes its parameters in response to traffic length in a lane, optimizing the road use and the signal timing of an intersection will benefit from being adapted to the dominant flows changing over the time of the day. The experiment analysis reveals that response time for green light in real-time increases in the lane with a greater number of vehicles and is decreased for the lane with lesser number of vehicles keeping the total time the same, so effective in managing traffic.
印度大城市的交通威胁导致许多旅客每天都在受苦。在交通控制中，迄今为止使用的是简单而古老的信号控制器，即机电信号控制器，它使用表盘计时器，具有固定的、有信号的交叉口时间计划。由于时间是固定的，车辆最多的车道上的人必须等待的时间最长，这导致了时间、金钱和汽油、柴油等自然资源的浪费。该系统是一个实时反馈的交通灯系统。通过摄像头检测特定车道上的车辆，然后使用深度学习算法YOLO (You Only Look Once)检测车道上的车辆总数，用于反馈控制车灯。交通灯控制器根据车道上的交通长度改变其参数，优化道路使用和十字路口的信号定时，将受益于适应一天中不同时间的主要流量变化。实验分析表明，在保持总时间不变的情况下，车辆数量较多的车道实时绿灯响应时间增大，车辆数量较少的车道实时绿灯响应时间减小，具有较好的交通管理效果。
{"title":"Automated Traffic Management System Using Deep Learning Based Object Detection","authors":"Sumindar Kaur Saini, Mankaran Singh Ghumman","doi":"10.1109/ICMLC56445.2022.9941332","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941332","url":null,"abstract":"The traffic menace in India’s metropolitan cities causes many travelers to suffer daily. In traffic control, simple and old forms of signal controllers, known as electro-mechanical signal controllers, are used till-date which use dial timers that have fixed, signalized intersection time plans. As the time is fixed, the people in the lane with the greatest number of vehicles must wait the most, leading to wastage of time, money, and natural resources such as petrol and diesel. The proposed system is a traffic light system with feedback in real-time. The vehicles present in a specific lane are detected using a camera and then the deep learning algorithm, YOLO (You Only Look Once) detects the total number of vehicles in a lane which is used for feedback control of the lights. The traffic lights controller changes its parameters in response to traffic length in a lane, optimizing the road use and the signal timing of an intersection will benefit from being adapted to the dominant flows changing over the time of the day. The experiment analysis reveals that response time for green light in real-time increases in the lane with a greater number of vehicles and is decreased for the lane with lesser number of vehicles keeping the total time the same, so effective in managing traffic.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130723504","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941311
Taichi Haba, Gaochao Cui, Hideaki Touyama
Brain-computer interface is mainly developed for clinical rehabilitation. Numerous studies have shown that it can also be applied to neuromarketing to assist customers in making decisions. By identifying the P300 component of the event-related potentials (ERPs), it can be known whether the target commodity or target stimuli is interesting to the consumer. However, when the target stimuli appear more frequently and people’s responses to stimuli vary, it is challenging to locate the target stimuli based on the P300 in practical applications. Moreover, a significant P300 component can only be obtained by stacking and averaging multiple ERPs in normal conditions. In this study, we propose a group electroencephalogram processing method to estimate the timing of evoked stimulus appearance without compromising real-time performance using convolutional neural networks. In addition, this method can be used to estimate the group’s attention to the target and standard stimulus. The results show that the effectiveness of the proposed processing method for stimuli presentation time estimation and group attention state estimation are 87.10 % and 96.55 %, respectively.
{"title":"Estimation of Stimulus Time and Average Attention State Based on Collective Addition of Event-Related Electroencephalography","authors":"Taichi Haba, Gaochao Cui, Hideaki Touyama","doi":"10.1109/ICMLC56445.2022.9941311","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941311","url":null,"abstract":"Brain-computer interface is mainly developed for clinical rehabilitation. Numerous studies have shown that it can also be applied to neuromarketing to assist customers in making decisions. By identifying the P300 component of the event-related potentials (ERPs), it can be known whether the target commodity or target stimuli is interesting to the consumer. However, when the target stimuli appear more frequently and people’s responses to stimuli vary, it is challenging to locate the target stimuli based on the P300 in practical applications. Moreover, a significant P300 component can only be obtained by stacking and averaging multiple ERPs in normal conditions. In this study, we propose a group electroencephalogram processing method to estimate the timing of evoked stimulus appearance without compromising real-time performance using convolutional neural networks. In addition, this method can be used to estimate the group’s attention to the target and standard stimulus. The results show that the effectiveness of the proposed processing method for stimuli presentation time estimation and group attention state estimation are 87.10 % and 96.55 %, respectively.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128027777","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 : 2022-09-09DOI: 10.1109/icmlc56445.2022.9941305
{"title":"ICMLC 2022 Cover Page","authors":"","doi":"10.1109/icmlc56445.2022.9941305","DOIUrl":"https://doi.org/10.1109/icmlc56445.2022.9941305","url":null,"abstract":"","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"86 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134529880","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941317
Kohei Hayashi, N. Yagi, Yutaka Hata, Yoshiaki Saji, Y. Sakai
Down syndrome is the most common chromosomal abnormality. In recent years, the frequency of births with Down syndrome has been increasing in Japan. One of the reasons for this is the trend toward late childbearing. Many children with Down syndrome often have gait problems such as Genu valgum and Genu varum. In order to solve such gait problems, custom- made insoles need to be created. This is because the shape of the feet and the way of walking vary from patient to patient. In addition, it is necessary to investigate whether these custom-made insoles are suitable for the patients with Down syndrome or not. Currently, the evaluation is done visually by doctors and physical therapists, however the criteria for judgment are unclear. Therefore, we worked to develop a system to determine whether custom-made insoles would improve Genu valgum and Genu varum. Since the children with Down syndrome rarely walk straight when taking gait videos, we focused on performing gait analysis in 3D instead of 2D, which has been performed previously. In this study, we estimated the joint position coordinates of a person from a walking video using 3D pose estimation in order to quantitatively evaluate the gait condition of children with Down syndrome. Additionally, by detecting the time of knee joint loading and measuring the knee joint angle, we were able to propose a system that can confirm symptoms such as genu valgum and genu varum. This system is expected to be easier than the existing 3D analyzers for gait analysis of patients with Down syndrome. As a future work, the accuracy of the system itself needs to be evaluated using an existing 3D motion analyzer that measures with markers.
{"title":"Time Series Knee Joint Angle Analysis During Gait for Patients with Down Syndrome by 3d Pose Estimation","authors":"Kohei Hayashi, N. Yagi, Yutaka Hata, Yoshiaki Saji, Y. Sakai","doi":"10.1109/ICMLC56445.2022.9941317","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941317","url":null,"abstract":"Down syndrome is the most common chromosomal abnormality. In recent years, the frequency of births with Down syndrome has been increasing in Japan. One of the reasons for this is the trend toward late childbearing. Many children with Down syndrome often have gait problems such as Genu valgum and Genu varum. In order to solve such gait problems, custom- made insoles need to be created. This is because the shape of the feet and the way of walking vary from patient to patient. In addition, it is necessary to investigate whether these custom-made insoles are suitable for the patients with Down syndrome or not. Currently, the evaluation is done visually by doctors and physical therapists, however the criteria for judgment are unclear. Therefore, we worked to develop a system to determine whether custom-made insoles would improve Genu valgum and Genu varum. Since the children with Down syndrome rarely walk straight when taking gait videos, we focused on performing gait analysis in 3D instead of 2D, which has been performed previously. In this study, we estimated the joint position coordinates of a person from a walking video using 3D pose estimation in order to quantitatively evaluate the gait condition of children with Down syndrome. Additionally, by detecting the time of knee joint loading and measuring the knee joint angle, we were able to propose a system that can confirm symptoms such as genu valgum and genu varum. This system is expected to be easier than the existing 3D analyzers for gait analysis of patients with Down syndrome. As a future work, the accuracy of the system itself needs to be evaluated using an existing 3D motion analyzer that measures with markers.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129619372","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941338
Yoshiki Asami, T. Motoyoshi, K. Sawai, H. Masuta, Noboru Takagi
This study presents a method for effectively applying formal concept analysis (FCA) to performance data for a practice-based Office E-learning system. Efforts to improve the content structure and design of an E-learning system typically involve the analysis of historical data; the problem is that the analyst generally selects the target of the analysis arbitrarily. We examined whether FCA can be used as a trigger for analysts to select the appropriate content. Specifically, we compare the implication relation between correct/incorrect questions captured by the implications of FCA and the overall trend obtained from statistical analysis methods.
{"title":"Examination of Analysis Methods for E-Learning System Grade Data Using Formal Concept Analysis","authors":"Yoshiki Asami, T. Motoyoshi, K. Sawai, H. Masuta, Noboru Takagi","doi":"10.1109/ICMLC56445.2022.9941338","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941338","url":null,"abstract":"This study presents a method for effectively applying formal concept analysis (FCA) to performance data for a practice-based Office E-learning system. Efforts to improve the content structure and design of an E-learning system typically involve the analysis of historical data; the problem is that the analyst generally selects the target of the analysis arbitrarily. We examined whether FCA can be used as a trigger for analysts to select the appropriate content. Specifically, we compare the implication relation between correct/incorrect questions captured by the implications of FCA and the overall trend obtained from statistical analysis methods.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"79 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116014647","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 : 2022-09-09DOI: 10.1109/ICMLC56445.2022.9941333
Kazunori Oka, Anas M. Ali, Daisuke Fujita, Syoji Kobashi
In the current dental practice, many panoramic dental images of the oral cavity are taken by x-ray radiograph. Using the dental panoramic images, a physician or dental assistant records dental chart. These burdens can deteriorate the quality of medical care, such as erroneous entries. Therefore, automatic analysis of panoramic dental images is desired. We have previously proposed a teeth recognition method based on Faster R-CNN and an optimization approach that performed a 94.2% accuracy. However, it shows a relatively low accuracy in panoramic images with prostheses. This paper proposed a new method to improve the accuracy by detecting prostheses separately. It first detects four types of prosthetic teeth using YOLOv5. Then, it recognizes the teeth and the prosthetic teeth simultaneously based on the proposed optimization approach using a prior knowledge model. The proposed method achieved a maximum recognition accuracy of 97.17%. It shows the usefulness of optimization using prior knowledge models in combination with prosthetic tooth detection.
{"title":"Tooth Recognition in X-Ray Dental Panoramic Images with Prosthetic Detection","authors":"Kazunori Oka, Anas M. Ali, Daisuke Fujita, Syoji Kobashi","doi":"10.1109/ICMLC56445.2022.9941333","DOIUrl":"https://doi.org/10.1109/ICMLC56445.2022.9941333","url":null,"abstract":"In the current dental practice, many panoramic dental images of the oral cavity are taken by x-ray radiograph. Using the dental panoramic images, a physician or dental assistant records dental chart. These burdens can deteriorate the quality of medical care, such as erroneous entries. Therefore, automatic analysis of panoramic dental images is desired. We have previously proposed a teeth recognition method based on Faster R-CNN and an optimization approach that performed a 94.2% accuracy. However, it shows a relatively low accuracy in panoramic images with prostheses. This paper proposed a new method to improve the accuracy by detecting prostheses separately. It first detects four types of prosthetic teeth using YOLOv5. Then, it recognizes the teeth and the prosthetic teeth simultaneously based on the proposed optimization approach using a prior knowledge model. The proposed method achieved a maximum recognition accuracy of 97.17%. It shows the usefulness of optimization using prior knowledge models in combination with prosthetic tooth detection.","PeriodicalId":117829,"journal":{"name":"2022 International Conference on Machine Learning and Cybernetics (ICMLC)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-09-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131170320","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：