Pub Date : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377360
Zhafri Hariz Roslan, Zalizah Awang Long, R. Ismail
The detection performance of tree crowns in forest environment has not been satisfactory compared to common objects, especially using aerial RGB imagery. Previous methods regarding Individual Tree Crown Detection (ITCD) utilizes different data sources to improve the detection rate due to the noisy image. Image enhancement methods such as super-resolution provide a solution to the noisy image by reconstructing the image using the low-resolution image. Generative Adversarial Network (GAN)-based model has shown success in super-resolution techniques. However, the GAN-based model created artefacts that may hinder the accuracy of the detection. In this paper, a noise-cancelling GAN-based model is proposed by averaging the weights of a compressed image and non-compressed image. The proposed method forces the network to discriminate the noise to generate a more photorealistic image. This method is inspired by super-resolution GAN (SRGAN) architecture with Residual Dense Network as the generator network. A two-stage object detection RetinaNet model is then used to detect the individual tree crowns in a sequential fashion. Extensive experiments have been conducted on a self-assembled tree crown dataset which showed the proposed model is more superior than a non-enhanced model with 0.6017 and 0.5908 respectively. Based on the results of the proposed method, the super-resolution technique can be used in conjunction with object detection algorithm to improve the detection in ITCD to improve the detection rate.
{"title":"Individual Tree Crown Detection using GAN and RetinaNet on Tropical Forest","authors":"Zhafri Hariz Roslan, Zalizah Awang Long, R. Ismail","doi":"10.1109/IMCOM51814.2021.9377360","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377360","url":null,"abstract":"The detection performance of tree crowns in forest environment has not been satisfactory compared to common objects, especially using aerial RGB imagery. Previous methods regarding Individual Tree Crown Detection (ITCD) utilizes different data sources to improve the detection rate due to the noisy image. Image enhancement methods such as super-resolution provide a solution to the noisy image by reconstructing the image using the low-resolution image. Generative Adversarial Network (GAN)-based model has shown success in super-resolution techniques. However, the GAN-based model created artefacts that may hinder the accuracy of the detection. In this paper, a noise-cancelling GAN-based model is proposed by averaging the weights of a compressed image and non-compressed image. The proposed method forces the network to discriminate the noise to generate a more photorealistic image. This method is inspired by super-resolution GAN (SRGAN) architecture with Residual Dense Network as the generator network. A two-stage object detection RetinaNet model is then used to detect the individual tree crowns in a sequential fashion. Extensive experiments have been conducted on a self-assembled tree crown dataset which showed the proposed model is more superior than a non-enhanced model with 0.6017 and 0.5908 respectively. Based on the results of the proposed method, the super-resolution technique can be used in conjunction with object detection algorithm to improve the detection in ITCD to improve the detection rate.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"39 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125207598","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377352
Thanh Cong Tran, T. K. Dang
Online transactions have increased drastically over the past decades. Credit card transactions account for a large percentage of these transactions. This leads to rise activities of credit card fraud transactions, causing losses in the finance industry. Therefore, it is vital to create reliable fraud detection systems, including two labels of fraud and no-fraud. However, there are highly unbalanced data between these two labels. In this paper, we use two resampling approaches of synthetic minority oversampling technique (SMOTE) and adaptive synthetic (ADASYN) to handle an imbalanced dataset to obtain the balanced dataset. The machine learning (ML) algorithms, named random forest, k nearest neighbors, decision tree, and logistic regression are applied to this balanced dataset. The comprehensive classification measurements, including fundamental, combined, and graphical measurements are used to evaluate the performances of these models. We observe that after resampling the dataset, the ML algorithms mentioned show the positive results of classification for fraudulent activities.
{"title":"Machine Learning for Prediction of Imbalanced Data: Credit Fraud Detection","authors":"Thanh Cong Tran, T. K. Dang","doi":"10.1109/IMCOM51814.2021.9377352","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377352","url":null,"abstract":"Online transactions have increased drastically over the past decades. Credit card transactions account for a large percentage of these transactions. This leads to rise activities of credit card fraud transactions, causing losses in the finance industry. Therefore, it is vital to create reliable fraud detection systems, including two labels of fraud and no-fraud. However, there are highly unbalanced data between these two labels. In this paper, we use two resampling approaches of synthetic minority oversampling technique (SMOTE) and adaptive synthetic (ADASYN) to handle an imbalanced dataset to obtain the balanced dataset. The machine learning (ML) algorithms, named random forest, k nearest neighbors, decision tree, and logistic regression are applied to this balanced dataset. The comprehensive classification measurements, including fundamental, combined, and graphical measurements are used to evaluate the performances of these models. We observe that after resampling the dataset, the ML algorithms mentioned show the positive results of classification for fraudulent activities.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117336835","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377363
Quang T. M. Pham, Jitae Shin
Age- related macular degeneration (AMD) is a disease of the central retina, which is one of the main reasons for vision loss of elderly people. To monitor the level of AMD, the doctors mainly use the retinal fundus images. However, the quality of retinal images can be affected during the imaging process. It leads to low contrast and blurry images. Those bad quality images can not be used for analyzing and diagnosis. For that reason, there are many studies about image enhancement in order to improve the quality of retinal photography. However, previous methods could not guarantee to keep the disease information after the enhancement process. Therefore, we introduce a generative adversarial model for AMD retinal image enhancement with additional factors to preserve the disease information. By exploiting drusen segmentation masks, our proposed model can enhance retinal photography quality and keep the pathological information.
{"title":"Generative Adversarial Networks for Retinal Image Enhancement with Pathological Information","authors":"Quang T. M. Pham, Jitae Shin","doi":"10.1109/IMCOM51814.2021.9377363","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377363","url":null,"abstract":"Age- related macular degeneration (AMD) is a disease of the central retina, which is one of the main reasons for vision loss of elderly people. To monitor the level of AMD, the doctors mainly use the retinal fundus images. However, the quality of retinal images can be affected during the imaging process. It leads to low contrast and blurry images. Those bad quality images can not be used for analyzing and diagnosis. For that reason, there are many studies about image enhancement in order to improve the quality of retinal photography. However, previous methods could not guarantee to keep the disease information after the enhancement process. Therefore, we introduce a generative adversarial model for AMD retinal image enhancement with additional factors to preserve the disease information. By exploiting drusen segmentation masks, our proposed model can enhance retinal photography quality and keep the pathological information.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133827362","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377430
Yuichi Chiken, D. Kitayama
When users perform a fashion coordination search, they tend to select their “preferred styles,” or styles that match their personal tastes, from among the search results. However, for effective fashion coordination, it is also important for one's fashion style to not only be suited to one's own preferences but also be highly evaluated by other people. In this paper, we propose a method that displays the evaluation values, such as the number of favorites, of fashion coordinates in the search results not when the search results are first displayed, but after the user has selected his/her preference from among the search results. The purpose of this method is to emphasize the difference between how a user perceives his/her own preference and how it is evaluated by other people, and to help users become aware of the differences between their preferred styles and others' preferences and perceptions. In this study, we implemented a system for fashion coordination search wherein the inputs are the desired items and styles from the user. In addition, we designed an experiment wherein we would vary the presentation timing of metadata such as user evaluation and style information in the search results. This experimentation method would clarify the effect of the system on changing users' attitudes toward fashion styling with respect to the timing of the presentation of user evaluation scores.
{"title":"Method for Changing Users' Attitudes Towards Fashion Styling by Showing Evaluations After Coordinate Selection","authors":"Yuichi Chiken, D. Kitayama","doi":"10.1109/IMCOM51814.2021.9377430","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377430","url":null,"abstract":"When users perform a fashion coordination search, they tend to select their “preferred styles,” or styles that match their personal tastes, from among the search results. However, for effective fashion coordination, it is also important for one's fashion style to not only be suited to one's own preferences but also be highly evaluated by other people. In this paper, we propose a method that displays the evaluation values, such as the number of favorites, of fashion coordinates in the search results not when the search results are first displayed, but after the user has selected his/her preference from among the search results. The purpose of this method is to emphasize the difference between how a user perceives his/her own preference and how it is evaluated by other people, and to help users become aware of the differences between their preferred styles and others' preferences and perceptions. In this study, we implemented a system for fashion coordination search wherein the inputs are the desired items and styles from the user. In addition, we designed an experiment wherein we would vary the presentation timing of metadata such as user evaluation and style information in the search results. This experimentation method would clarify the effect of the system on changing users' attitudes toward fashion styling with respect to the timing of the presentation of user evaluation scores.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"29 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"114462931","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377415
Li-Hua Li, Chang-Yu Lai
In today's rising consumer awareness, companies are paying more and more attention to customer satisfaction. In order to survive in a fiercely competitive environment and maintain their competitive advantages, the only way to continuously provide consumers with high-quality services is the key to the sustainable operation of modern enterprises. The purpose of this research is focusing on the impact of service quality for automotive aftermarket parts and customers' willingness to repurchase. In this study, 400 questionnaire invitations through e-mail were distributed to existing customers and 164 valid questionnaires were responded. The responded answers were encoded, filed, and verified using SPSS. Degree and validity analysis, narrative statistics, single factor analysis of variance (ANOVA), regression analysis and structural equation modeling were applied for analysis. Through empirical analysis, there are many findings: Sales Service & Marketing, R&D capabilities, and innovative services in service quality are positively and significantly related to customers' willingness to repurchase. In the single factor variation analysis and structural equations, it is found that the impact of customer type on service quality and customer repurchase intention is not significantly related. In this study, Artificial Intelligence (AI) was also applied to compare the impact of service quality and to build the prediction model for customer repurchasing. These AI techniques include decision tree, neural network models, and multiple-linear regression. It is concluded that Artificial Neural Networks (ANN) have better predictive ability after training with sufficient data and proper input data. For decision tree and regression analysis, these models' predicting power will decrease when the data becomes more complex.
{"title":"Comparing the Impact of Service Quality on Customers' Repurchase Intentions Based on Statistical Methods and Artificial Intelligence-Taking an Automotive Aftermarket (AM) Parts Sales Company as an Example","authors":"Li-Hua Li, Chang-Yu Lai","doi":"10.1109/IMCOM51814.2021.9377415","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377415","url":null,"abstract":"In today's rising consumer awareness, companies are paying more and more attention to customer satisfaction. In order to survive in a fiercely competitive environment and maintain their competitive advantages, the only way to continuously provide consumers with high-quality services is the key to the sustainable operation of modern enterprises. The purpose of this research is focusing on the impact of service quality for automotive aftermarket parts and customers' willingness to repurchase. In this study, 400 questionnaire invitations through e-mail were distributed to existing customers and 164 valid questionnaires were responded. The responded answers were encoded, filed, and verified using SPSS. Degree and validity analysis, narrative statistics, single factor analysis of variance (ANOVA), regression analysis and structural equation modeling were applied for analysis. Through empirical analysis, there are many findings: Sales Service & Marketing, R&D capabilities, and innovative services in service quality are positively and significantly related to customers' willingness to repurchase. In the single factor variation analysis and structural equations, it is found that the impact of customer type on service quality and customer repurchase intention is not significantly related. In this study, Artificial Intelligence (AI) was also applied to compare the impact of service quality and to build the prediction model for customer repurchasing. These AI techniques include decision tree, neural network models, and multiple-linear regression. It is concluded that Artificial Neural Networks (ANN) have better predictive ability after training with sufficient data and proper input data. For decision tree and regression analysis, these models' predicting power will decrease when the data becomes more complex.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124813142","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377387
Hyewon Kim, H. Jeon
Although the attempts to use virtual reality (VR) for stress reduction or relaxation are increasing, the evidence on who will benefit is still lacking. In this study, we aimed to identify the clinical and physiological predictors for effectiveness of stress reduction or relaxation using VR. 83 healthy, but highly stressed adults were enrolled for the study. At baseline, demographic information and medical history were collected and physiological parameters including heart rate variability were extracted. Subjects were evaluated subjective discomfort using the State-Trait Anxiety Inventory-X-1, the 0–100 Numeric rating scale repetitively throughout the VR application. To identify the predictors for the effectiveness of VR relaxation, correlation analyses and multivariate regression analyses were conducted. As results, we found that smoking is negatively associated with the effectiveness of VR relaxation and baseline subjective discomfort, respiratory rate and heart rate are positively associated with the effectiveness of VR relaxation. This suggest that the effect of VR relaxation is large in people with high respiratory rate and heart rate, and that the effect is reduced in smokers.
尽管使用虚拟现实(VR)来减压或放松的尝试越来越多，但谁将从中受益的证据仍然缺乏。在这项研究中，我们旨在确定使用VR减压或放松效果的临床和生理预测因素。83名健康但压力很大的成年人参加了这项研究。在基线时，收集人口统计信息和病史，并提取包括心率变异性在内的生理参数。在整个虚拟现实应用过程中，使用状态-特质焦虑量表- x -1(0-100数值评定量表)对受试者进行主观不适评估。为了确定VR放松效果的预测因素，进行了相关分析和多元回归分析。结果，我们发现吸烟与VR放松的有效性呈负相关，而基线主观不适感、呼吸频率和心率与VR放松的有效性呈正相关。这表明，VR放松对呼吸频率和心率高的人的影响很大，而对吸烟者的影响较小。
{"title":"Predicting who will benefit from relaxation or stress reduction through virtual reality","authors":"Hyewon Kim, H. Jeon","doi":"10.1109/IMCOM51814.2021.9377387","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377387","url":null,"abstract":"Although the attempts to use virtual reality (VR) for stress reduction or relaxation are increasing, the evidence on who will benefit is still lacking. In this study, we aimed to identify the clinical and physiological predictors for effectiveness of stress reduction or relaxation using VR. 83 healthy, but highly stressed adults were enrolled for the study. At baseline, demographic information and medical history were collected and physiological parameters including heart rate variability were extracted. Subjects were evaluated subjective discomfort using the State-Trait Anxiety Inventory-X-1, the 0–100 Numeric rating scale repetitively throughout the VR application. To identify the predictors for the effectiveness of VR relaxation, correlation analyses and multivariate regression analyses were conducted. As results, we found that smoking is negatively associated with the effectiveness of VR relaxation and baseline subjective discomfort, respiratory rate and heart rate are positively associated with the effectiveness of VR relaxation. This suggest that the effect of VR relaxation is large in people with high respiratory rate and heart rate, and that the effect is reduced in smokers.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"28 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121926187","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}
The goal of networking has the idea of “resource sharing” and “communication” in a convenient way. However, more convenience services are provided, more problems of security and privacy issues may occur. In order to prevent these problems, an IDS (Intrusion Detection System) is designed to enhance the network security and to observe abnormal behavior. Model accuracy and the training time required to build the model are affected greatly if we use the unselected features and irrelevant data. This is the reason why the selection of features is a significant process in building an Intrusion Detection System (IDS). This paper aims to boost the Deep Neural Network (DNN) capabilities by selecting the feasible features before processing networking data. This research employed the KDD Cup 99 dataset which is considered as one of the representative datasets for intrusion detection. Based on our experimental results, it is concluded that the selection of the proper features has effects on the improvement of IDS compared to the method without feature selection. This research has proved that the improvement of DNN for IDS can reach up to 99.4% for accuracy, 99.7% for precision, 97.9% for recall, and 98.8 for F1 score.
网络化的目标是以方便的方式实现“资源共享”和“交流”。然而，在提供更多便利服务的同时，也可能出现更多的安全和隐私问题。为了防止这些问题的发生，我们设计了入侵检测系统(IDS, Intrusion Detection System)来增强网络的安全性并观察异常行为。如果我们使用未选择的特征和不相关的数据，模型的准确性和建立模型所需的训练时间都会受到很大的影响。这就是为什么特征选择是构建入侵检测系统(IDS)的一个重要过程。本文旨在通过在处理网络数据之前选择可行的特征来提高深度神经网络(DNN)的能力。本研究采用了被认为是入侵检测的代表性数据集之一的KDD Cup 99数据集。实验结果表明，与不选择特征的方法相比，选择合适的特征对IDS的改进有一定的影响。本研究证明，DNN对IDS的准确率提高了99.4%，准确率提高了99.7%，召回率提高了97.9%，F1分数提高了98.8。
{"title":"A Feature Selection Based DNN for Intrusion Detection System","authors":"Li-Hua Li, Ramli Ahmad, Weng-Chung Tsai, Alok Kumar Sharma","doi":"10.1109/IMCOM51814.2021.9377405","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377405","url":null,"abstract":"The goal of networking has the idea of “resource sharing” and “communication” in a convenient way. However, more convenience services are provided, more problems of security and privacy issues may occur. In order to prevent these problems, an IDS (Intrusion Detection System) is designed to enhance the network security and to observe abnormal behavior. Model accuracy and the training time required to build the model are affected greatly if we use the unselected features and irrelevant data. This is the reason why the selection of features is a significant process in building an Intrusion Detection System (IDS). This paper aims to boost the Deep Neural Network (DNN) capabilities by selecting the feasible features before processing networking data. This research employed the KDD Cup 99 dataset which is considered as one of the representative datasets for intrusion detection. Based on our experimental results, it is concluded that the selection of the proper features has effects on the improvement of IDS compared to the method without feature selection. This research has proved that the improvement of DNN for IDS can reach up to 99.4% for accuracy, 99.7% for precision, 97.9% for recall, and 98.8 for F1 score.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127570361","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377413
Arefeh Yavary, H. Sajedi
Document Embedding methods are an impressive task in each machine learning or neural network based natural language processing task. This task is entitled by representation learning and knowledge representation, too. In ultimate the target of this task, each document outputs a representation format of text documents in order to be understandable for machine. Literature reviews in representation learning, shows that document embedding methods for text is weaker in compare with representation of image or signal. Also, in compare to other data like as image or signal, representation of text has more challenges. By this, this paper we suggested a piped process of Generative Adversarial Neural Network and Extreme Learning Machine technique for document embedding. The experimental results show that document embedding using this combination of Generative Adversarial Networks and Extreme learning machines is comparative with other available methods of document embedding.
{"title":"Document Embedding using piped ELM-GAN Model","authors":"Arefeh Yavary, H. Sajedi","doi":"10.1109/IMCOM51814.2021.9377413","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377413","url":null,"abstract":"Document Embedding methods are an impressive task in each machine learning or neural network based natural language processing task. This task is entitled by representation learning and knowledge representation, too. In ultimate the target of this task, each document outputs a representation format of text documents in order to be understandable for machine. Literature reviews in representation learning, shows that document embedding methods for text is weaker in compare with representation of image or signal. Also, in compare to other data like as image or signal, representation of text has more challenges. By this, this paper we suggested a piped process of Generative Adversarial Neural Network and Extreme Learning Machine technique for document embedding. The experimental results show that document embedding using this combination of Generative Adversarial Networks and Extreme learning machines is comparative with other available methods of document embedding.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"41 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129887366","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377410
Md. Tanvir Hossain, Md. Wahid Hasan, A. Das
In recent years, Machine Learning and Data Mining based research become prevalent and handwritten recognition is one of the hotcakes. Bangla handwritten word recognition and extraction acquired huge attention in many research sectors like Computer Vision, Image Processing, Machine Learning, and many others for a large field of applications. To tackle this challenging problem, a perfect segmentation and recognition method are described in this paper with a good percentage of accuracy. The main challenge was to introduce a sound segmentation system and merge multi-zoned characters. This paper proposes a multi-zoned character segmentation, and a merging method is also proposed, which can produce the handwritten term. Utilizing Convolutional Neural Network (CNN) for preparing 84% precision is accomplished for character level, and 82% precision is achieved in word level.
{"title":"Bangla Handwritten Word Recognition System Using Convolutional Neural Network","authors":"Md. Tanvir Hossain, Md. Wahid Hasan, A. Das","doi":"10.1109/IMCOM51814.2021.9377410","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377410","url":null,"abstract":"In recent years, Machine Learning and Data Mining based research become prevalent and handwritten recognition is one of the hotcakes. Bangla handwritten word recognition and extraction acquired huge attention in many research sectors like Computer Vision, Image Processing, Machine Learning, and many others for a large field of applications. To tackle this challenging problem, a perfect segmentation and recognition method are described in this paper with a good percentage of accuracy. The main challenge was to introduce a sound segmentation system and merge multi-zoned characters. This paper proposes a multi-zoned character segmentation, and a merging method is also proposed, which can produce the handwritten term. Utilizing Convolutional Neural Network (CNN) for preparing 84% precision is accomplished for character level, and 82% precision is achieved in word level.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127863394","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 : 2021-01-04DOI: 10.1109/IMCOM51814.2021.9377435
Tri Trinh, N. V. Bang, Manabu Yoshino, H. Suzuki, Nguyen Huu Thanh
We propose a novel server design implementing Fiber-to-the-Home (FTTH) Zero-Touch Multi-service Provisioning (ZTMP) technology, which is gaining significant attention from telephone companies (Telcos) specifically the Vietnam Post and Telecommunications Group because it can minimize human errors and labor costs. Vendor FTTH ZTMP solutions are often considered ossified and not suitable for implementation into Telco auto-provisioning workflows, which are complicated, Telco specific, and subject to change over time. The proposed ZTMP server design incorporates an open source platform approach based on a Central Office Re-architected as a Data Center (CORD). We deploy the proposed server in an Asia-Pacific CORD-based FTTH ZTMP test bed system. Evaluation results show that the system-control flow functions well in the test-bed.
{"title":"FTTH Zero-Touch Multi-Service Provisioning on CORD-Based Access Network Virtualization Platform","authors":"Tri Trinh, N. V. Bang, Manabu Yoshino, H. Suzuki, Nguyen Huu Thanh","doi":"10.1109/IMCOM51814.2021.9377435","DOIUrl":"https://doi.org/10.1109/IMCOM51814.2021.9377435","url":null,"abstract":"We propose a novel server design implementing Fiber-to-the-Home (FTTH) Zero-Touch Multi-service Provisioning (ZTMP) technology, which is gaining significant attention from telephone companies (Telcos) specifically the Vietnam Post and Telecommunications Group because it can minimize human errors and labor costs. Vendor FTTH ZTMP solutions are often considered ossified and not suitable for implementation into Telco auto-provisioning workflows, which are complicated, Telco specific, and subject to change over time. The proposed ZTMP server design incorporates an open source platform approach based on a Central Office Re-architected as a Data Center (CORD). We deploy the proposed server in an Asia-Pacific CORD-based FTTH ZTMP test bed system. Evaluation results show that the system-control flow functions well in the test-bed.","PeriodicalId":275121,"journal":{"name":"2021 15th International Conference on Ubiquitous Information Management and Communication (IMCOM)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-01-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126614242","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：