Pub Date : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932996
George Gkolfopoulos, Iraklis Varlamis
Text categorization is a significant task in the re-search field of text mining, which has recently benefited from deep neural network algorithms and advanced learning techniques that extract language models from large textual corpora. These Pre-Trained Language Models are the main components of state-of-the-art solutions in many natural language processing and text-mining tasks can be very generic, trained in generic text corpora, or domain-specific when they employ large corpora from specific application domains (e.g. social media, news, sciences, etc.). When only generic language models are available the overall performance in the task can be improved by adapting or fine-tuning the model used for the task, e.g. the classifier. Although multilingual language models are reported in the literature, such models are usually language-specific. This work presents a news article classifier, which has been trained on a small corpus and employs a Greek version of BERT language model. Comparison with existing machine learning-based classifiers shows that the proposed method outperforms well-known methods in text classification. In addition, the proposed approach allows the continuous training of the classifier through user-provided feedback on falsely classified articles.
{"title":"Developing a news classifier for Greek using BERT","authors":"George Gkolfopoulos, Iraklis Varlamis","doi":"10.1109/SEEDA-CECNSM57760.2022.9932996","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932996","url":null,"abstract":"Text categorization is a significant task in the re-search field of text mining, which has recently benefited from deep neural network algorithms and advanced learning techniques that extract language models from large textual corpora. These Pre-Trained Language Models are the main components of state-of-the-art solutions in many natural language processing and text-mining tasks can be very generic, trained in generic text corpora, or domain-specific when they employ large corpora from specific application domains (e.g. social media, news, sciences, etc.). When only generic language models are available the overall performance in the task can be improved by adapting or fine-tuning the model used for the task, e.g. the classifier. Although multilingual language models are reported in the literature, such models are usually language-specific. This work presents a news article classifier, which has been trained on a small corpus and employs a Greek version of BERT language model. Comparison with existing machine learning-based classifiers shows that the proposed method outperforms well-known methods in text classification. In addition, the proposed approach allows the continuous training of the classifier through user-provided feedback on falsely classified articles.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77510273","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932927
Giannis Botilias, Lamprini Pappa, P. Karvelis, C. Stylios
The widespread availability of smartphones and their high processing power have made them powerful mobile tools able to host and run various apps. In addition, wearable devices with low cost and accurate sensors gathering various physiological data and information are now available. Meanwhile, automated activity recognition is a rapidly evolving research area directly related to the mobile Health (mHealth) field. Rapid advancements in the Human Activity Recognition (HAR) field are mainly based on combining smartphones and wearable devices to succeed in advancing health tracking. This paper presents a mobile app designed and developed for monitoring changes in variables related to the physiological health status of an individual when he is moving around. The app tracks the physiological status of a human along with machine learning algorithms able to recognize and identify human activity and produce automatic alerts warning of dangerous health situations.
{"title":"Tracking individuals’ health using mobile applications and Machine Learning","authors":"Giannis Botilias, Lamprini Pappa, P. Karvelis, C. Stylios","doi":"10.1109/SEEDA-CECNSM57760.2022.9932927","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932927","url":null,"abstract":"The widespread availability of smartphones and their high processing power have made them powerful mobile tools able to host and run various apps. In addition, wearable devices with low cost and accurate sensors gathering various physiological data and information are now available. Meanwhile, automated activity recognition is a rapidly evolving research area directly related to the mobile Health (mHealth) field. Rapid advancements in the Human Activity Recognition (HAR) field are mainly based on combining smartphones and wearable devices to succeed in advancing health tracking. This paper presents a mobile app designed and developed for monitoring changes in variables related to the physiological health status of an individual when he is moving around. The app tracks the physiological status of a human along with machine learning algorithms able to recognize and identify human activity and produce automatic alerts warning of dangerous health situations.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87751823","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9933002
N. Ploskas
Linear programming solvers include various options that can be used to control algorithmic aspects and considerably impact the solver performance. As it is obvious, manually finding optimal parameters is a very difficult task and sometimes impossible. For this reason, it is necessary to implement smart techniques that will automate this process. Other works have utilized derivative-free optimization solvers to tune solver parameters. In this work, eight open-source derivative-free optimization solvers are utilized for finding (near) optimal tuning parameters of state-of-the-art linear programming solvers. We investigate how sensitive linear programming solvers are to a parameter tuning process. Extensive computational results are presented on tuning four linear programming solvers (CLP, CPLEX, GUROBI, and XPRESS) over a set of 70 benchmark problems. We find better parameters for all linear programming solvers, achieving a reduction in execution time over their default parameters up to 26%. We conclude that several derivative-free optimization solvers outperform others on finding optimal optimal tuning parameters for linear programming solvers.
{"title":"Parameter Tuning of Linear Programming Solvers","authors":"N. Ploskas","doi":"10.1109/SEEDA-CECNSM57760.2022.9933002","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9933002","url":null,"abstract":"Linear programming solvers include various options that can be used to control algorithmic aspects and considerably impact the solver performance. As it is obvious, manually finding optimal parameters is a very difficult task and sometimes impossible. For this reason, it is necessary to implement smart techniques that will automate this process. Other works have utilized derivative-free optimization solvers to tune solver parameters. In this work, eight open-source derivative-free optimization solvers are utilized for finding (near) optimal tuning parameters of state-of-the-art linear programming solvers. We investigate how sensitive linear programming solvers are to a parameter tuning process. Extensive computational results are presented on tuning four linear programming solvers (CLP, CPLEX, GUROBI, and XPRESS) over a set of 70 benchmark problems. We find better parameters for all linear programming solvers, achieving a reduction in execution time over their default parameters up to 26%. We conclude that several derivative-free optimization solvers outperform others on finding optimal optimal tuning parameters for linear programming solvers.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74669836","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932936
Foteini Tatsi, Fotios Tatsis, K. Karamanis
The motives and factors influencing an individual’s decision to participate in the accommodation sharing economy differ from person to person. Profit and social interaction are highlighted as the main motives driving an individual to become a host. The number of hosts who manage more than one listing is constantly increasing. Managing more listings makes hosts more experienced in serving guests and making more profit. This paper introduces “beHost”, a web application built to guide potential first time hosts to become more competitive and achieve their goals.
{"title":"beHost: A web application for potential accommodation providers interested in participating in the sharing economy","authors":"Foteini Tatsi, Fotios Tatsis, K. Karamanis","doi":"10.1109/SEEDA-CECNSM57760.2022.9932936","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932936","url":null,"abstract":"The motives and factors influencing an individual’s decision to participate in the accommodation sharing economy differ from person to person. Profit and social interaction are highlighted as the main motives driving an individual to become a host. The number of hosts who manage more than one listing is constantly increasing. Managing more listings makes hosts more experienced in serving guests and making more profit. This paper introduces “beHost”, a web application built to guide potential first time hosts to become more competitive and achieve their goals.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79653929","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932957
V. Liagkou, S. Sakka, C. Stylios
Human activity recognition systems (HARS) should allow the secure and trustworthy exchange of sensitive data between several kinds of participating parties with different aims and claims, regarding security, data protection, and trust issues. Initially in this work, a security flaw has been identified in a complete medical IoT application using wearable devices and smart sensors. Then, we list the security vulnerabilities and attempt to make suggestions on the prevention of security flaws that may appear during the implementation of HARS and we analyze a specific attack, the Man in the Middle attack, where a third malicious entity interferes with communication between two entities and is associated with key exchange protocols. Moreover, we discuss various design considerations for protecting the data that is transmitted and stored from different sources like smart wearables, mobile phones, and cloud applications by using cryptographic and privacy-preserving techniques. Finally, we show how the use of the OAuth2.0 protocol can ensure that only authenticated users interact with the HARS.
{"title":"Security and Privacy Vulnerabilities in Human Activity Recognition systems","authors":"V. Liagkou, S. Sakka, C. Stylios","doi":"10.1109/SEEDA-CECNSM57760.2022.9932957","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932957","url":null,"abstract":"Human activity recognition systems (HARS) should allow the secure and trustworthy exchange of sensitive data between several kinds of participating parties with different aims and claims, regarding security, data protection, and trust issues. Initially in this work, a security flaw has been identified in a complete medical IoT application using wearable devices and smart sensors. Then, we list the security vulnerabilities and attempt to make suggestions on the prevention of security flaws that may appear during the implementation of HARS and we analyze a specific attack, the Man in the Middle attack, where a third malicious entity interferes with communication between two entities and is associated with key exchange protocols. Moreover, we discuss various design considerations for protecting the data that is transmitted and stored from different sources like smart wearables, mobile phones, and cloud applications by using cryptographic and privacy-preserving techniques. Finally, we show how the use of the OAuth2.0 protocol can ensure that only authenticated users interact with the HARS.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75453317","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932987
Andreas Miltiadous, Vasileios Aspiotis, Konstantinos Sakkas, N. Giannakeas, E. Glavas, A. Tzallas
Stress is a subject always relevant to scientific research due to the numerous implications in human life. Typical biomarkers used in the physiological evaluation of stress include Electrocardiography, cortisol levels, galvanic skin response and other. Recently, one less widely used instrument for the assessment of stress that has been re-emerged due to advancements in computational power and machine learning techniques, is Electroencephalography. Moreover, as Virtual Reality HMDs are being rapidly adopted by the research community it becomes apparent that leveraging the offered advantages of VR for the exploration of stress can lead to novel controlable and reproducable experimental procedures. In this paper we combine EEG, ECG and the Perceived Stress Scale with a Virtual Reality phobia induction setting, to propose a protocol for assessing stress. The suggested protocol can be used for functional brain connectivity investigation and thus the evaluation of stress while it and can be expanded via the incorporation of machine learning algorithms for automatic stress level classification.
{"title":"An experimental protocol for exploration of stress in an immersive VR scenario with EEG","authors":"Andreas Miltiadous, Vasileios Aspiotis, Konstantinos Sakkas, N. Giannakeas, E. Glavas, A. Tzallas","doi":"10.1109/SEEDA-CECNSM57760.2022.9932987","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932987","url":null,"abstract":"Stress is a subject always relevant to scientific research due to the numerous implications in human life. Typical biomarkers used in the physiological evaluation of stress include Electrocardiography, cortisol levels, galvanic skin response and other. Recently, one less widely used instrument for the assessment of stress that has been re-emerged due to advancements in computational power and machine learning techniques, is Electroencephalography. Moreover, as Virtual Reality HMDs are being rapidly adopted by the research community it becomes apparent that leveraging the offered advantages of VR for the exploration of stress can lead to novel controlable and reproducable experimental procedures. In this paper we combine EEG, ECG and the Perceived Stress Scale with a Virtual Reality phobia induction setting, to propose a protocol for assessing stress. The suggested protocol can be used for functional brain connectivity investigation and thus the evaluation of stress while it and can be expanded via the incorporation of machine learning algorithms for automatic stress level classification.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86464586","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932959
D. Amanatidis, Georgios Chatzisavvas, Michael F. Dossis
The diagnosis of an autoimmune disease usually requires a careful examination of the patient’s health history and the evaluation of any possible occupation and environment related exposures. Frequently, autoimmune disorders have early symptoms such as joint and muscle pain, fatigue, weight loss or fever. These symptoms however are non-specific and imaging technology tools can be extremely valuable for precise diagnosis. In this paper, we deal with autoimmune diseases that result in brain damage and more specifically, multiple sclerosis. Classification of brain MRI images is performed leveraging a Convolutional Neural Network, showing excellent results.
{"title":"Brain MRI based diagnosis of autoimmune diseases using deep learning","authors":"D. Amanatidis, Georgios Chatzisavvas, Michael F. Dossis","doi":"10.1109/SEEDA-CECNSM57760.2022.9932959","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932959","url":null,"abstract":"The diagnosis of an autoimmune disease usually requires a careful examination of the patient’s health history and the evaluation of any possible occupation and environment related exposures. Frequently, autoimmune disorders have early symptoms such as joint and muscle pain, fatigue, weight loss or fever. These symptoms however are non-specific and imaging technology tools can be extremely valuable for precise diagnosis. In this paper, we deal with autoimmune diseases that result in brain damage and more specifically, multiple sclerosis. Classification of brain MRI images is performed leveraging a Convolutional Neural Network, showing excellent results.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88982936","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932990
Christos N. Karras, Aristeidis Karras, D. Tsolis, K. Giotopoulos, S. Sioutas
Big data management methods are paramount in the modern era as applications tend to create massive amounts of data that comes from various sources. Therefore, there is an urge to create adaptive, speedy and robust frameworks that can effectively handle massive datasets. Distributed environments such as Apache Spark are of note, as they can handle such data by creating clusters where a portion of the data is stored locally and then the results are returned with the use of Resilient Distributed Datasets (RDDs). In this paper a method for distributed marginal Gibbs sampling for widely used latent Dirichlet allocation (LDA) model is implemented on PySpark along with a Metropolis Hastings Random Walker. The Distributed LDA (DLDA) algorithm distributes a given dataset into P partitions and performs local LDA on each partition, for each document independently. Every nth iteration, local LDA models, that were trained on distinct partitions, are combined to assure the model ability to converge. Experimental results are promising as the proposed system demonstrates comparable performance in the final model quality to the sequential LDA, and achieves significant speedup time-optimizations when utilized with massive datasets.
大数据管理方法在现代时代是至关重要的，因为应用程序往往会创建来自各种来源的大量数据。因此，迫切需要创建能够有效处理大量数据集的自适应、快速和健壮的框架。像Apache Spark这样的分布式环境是值得注意的，因为它们可以通过创建集群来处理这些数据，其中一部分数据存储在本地，然后使用弹性分布式数据集(rdd)返回结果。本文利用Metropolis Hastings Random Walker在PySpark上实现了广泛应用的潜在狄利克雷分配(latent Dirichlet allocation, LDA)模型的分布式边际Gibbs抽样方法。分布式LDA (Distributed LDA)算法将给定的数据集分布到P个分区中，并在每个分区上独立地对每个文档执行本地LDA。在每第n次迭代中，对在不同分区上训练的局部LDA模型进行组合，以确保模型的收敛能力。实验结果是有希望的，因为所提出的系统在最终模型质量方面表现出与顺序LDA相当的性能，并且在使用大量数据集时实现了显着的加速时间优化。
{"title":"Distributed Gibbs Sampling and LDA Modelling for Large Scale Big Data Management on PySpark","authors":"Christos N. Karras, Aristeidis Karras, D. Tsolis, K. Giotopoulos, S. Sioutas","doi":"10.1109/SEEDA-CECNSM57760.2022.9932990","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932990","url":null,"abstract":"Big data management methods are paramount in the modern era as applications tend to create massive amounts of data that comes from various sources. Therefore, there is an urge to create adaptive, speedy and robust frameworks that can effectively handle massive datasets. Distributed environments such as Apache Spark are of note, as they can handle such data by creating clusters where a portion of the data is stored locally and then the results are returned with the use of Resilient Distributed Datasets (RDDs). In this paper a method for distributed marginal Gibbs sampling for widely used latent Dirichlet allocation (LDA) model is implemented on PySpark along with a Metropolis Hastings Random Walker. The Distributed LDA (DLDA) algorithm distributes a given dataset into P partitions and performs local LDA on each partition, for each document independently. Every nth iteration, local LDA models, that were trained on distinct partitions, are combined to assure the model ability to converge. Experimental results are promising as the proposed system demonstrates comparable performance in the final model quality to the sequential LDA, and achieves significant speedup time-optimizations when utilized with massive datasets.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"88663874","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932913
Thomai Karamitsou, Dimitrios Seventekidis, Christos Karapiperis, Konstantina Banti, Ioanna Karampelia, Thomas S. Kyriakidis, M. Louta
Treated wastewater reuse is increasingly important for efficient and sustainable management of water resources due to increased water demands. Motivated by the above, AUGEIAS proposes an Internet of Things (IoT) approach for clean and treated wastewater usage in precision agriculture. In this context, real-time measurements for wastewater treatment plant and field are correlated with open data to improve crop water needs prediction mechanisms. This paper presents the open weather sources that are used and evaluates their reliability. After the open data is evaluated, it is integrated with the data collected by IoT sensors/devices. By using the mean absolute percentage error metric, we evaluate the forecasting performance of open weather sources. According to our study, OpenWeatherMap’s forecast data proved more accurate, with a success rate at 83.3%.
{"title":"Open weather data evaluation for crop irrigation prediction mechanisms in the AUGEIAS project","authors":"Thomai Karamitsou, Dimitrios Seventekidis, Christos Karapiperis, Konstantina Banti, Ioanna Karampelia, Thomas S. Kyriakidis, M. Louta","doi":"10.1109/SEEDA-CECNSM57760.2022.9932913","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932913","url":null,"abstract":"Treated wastewater reuse is increasingly important for efficient and sustainable management of water resources due to increased water demands. Motivated by the above, AUGEIAS proposes an Internet of Things (IoT) approach for clean and treated wastewater usage in precision agriculture. In this context, real-time measurements for wastewater treatment plant and field are correlated with open data to improve crop water needs prediction mechanisms. This paper presents the open weather sources that are used and evaluates their reliability. After the open data is evaluated, it is integrated with the data collected by IoT sensors/devices. By using the mean absolute percentage error metric, we evaluate the forecasting performance of open weather sources. According to our study, OpenWeatherMap’s forecast data proved more accurate, with a success rate at 83.3%.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"72385158","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 : 2022-09-23DOI: 10.1109/SEEDA-CECNSM57760.2022.9932921
Dimitrios Mpouziotas, Eleftherios Mastrapas, Nikos Dimokas, Petros Karvelis, E. Glavas
Object detection is a computer vision method for locating objects in images. Although, it has surpassed human performance and it has been considered practically solved, there are still considerable challenges, such as when photos are captured under suboptimal lighting conditions due to environmental and/or technical constraints. On the other hand, a variety of methods have been developed to enhance low light images, which can boost an object detector’s performance. In this work, we apply different image enhancement methods and study how they affect the efficacy of a well known detector (You Only Look Once, YOLO). A statistical analysis between YOLO’s performance for each enhancing algorithm, using a low light imaging dataset, is also presented, proving that for these kind of images, enhancement is a valuable step.
目标检测是一种定位图像中目标的计算机视觉方法。虽然，它已经超越了人类的表现，并被认为实际上已经解决了，但仍然存在相当大的挑战，例如由于环境和/或技术限制，当照片在次优照明条件下拍摄时。另一方面，已经开发了各种方法来增强低光图像，这可以提高目标检测器的性能。在这项工作中，我们应用了不同的图像增强方法，并研究了它们如何影响一个众所周知的检测器(You Only Look Once, YOLO)的有效性。在微光成像数据集上，对不同增强算法的YOLO性能进行了统计分析，证明了对这类图像进行增强是有价值的一步。
{"title":"Object Detection for Low Light Images","authors":"Dimitrios Mpouziotas, Eleftherios Mastrapas, Nikos Dimokas, Petros Karvelis, E. Glavas","doi":"10.1109/SEEDA-CECNSM57760.2022.9932921","DOIUrl":"https://doi.org/10.1109/SEEDA-CECNSM57760.2022.9932921","url":null,"abstract":"Object detection is a computer vision method for locating objects in images. Although, it has surpassed human performance and it has been considered practically solved, there are still considerable challenges, such as when photos are captured under suboptimal lighting conditions due to environmental and/or technical constraints. On the other hand, a variety of methods have been developed to enhance low light images, which can boost an object detector’s performance. In this work, we apply different image enhancement methods and study how they affect the efficacy of a well known detector (You Only Look Once, YOLO). A statistical analysis between YOLO’s performance for each enhancing algorithm, using a low light imaging dataset, is also presented, proving that for these kind of images, enhancement is a valuable step.","PeriodicalId":68279,"journal":{"name":"计算机工程与设计","volume":null,"pages":null},"PeriodicalIF":0.0,"publicationDate":"2022-09-23","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"83490922","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：