Pub Date : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751311
Ivan Jovović, Dejan Babic, Stevan Cakic, Tomo Popović, S. Krco, Petar Knezevic
This paper describes research effort aimed at the use of machine learning, Internet of Things, and edge computing for a use case in health, mainly the prevention of the spread of infectious diseases. The main motivation for the research was the Covid-19 pandemic and the need to improve control of the prevention measures implementation. In the study, the experimentation was focused on the use of machine learning to create and utilize prediction models for face mask detection. The prediction model is then evaluated on the various platforms with a focus on the use on various edge devices equipped with a video camera sensor. Different platforms have been tested and evaluated such as standard laptop PC, Raspberry Pi3, and Jetson Nano AI edge platform. Finally, the paper discusses a possible approach to implement a solution that would utilize the face mask detection function and lays out the path for the future research steps.
本文描述了旨在将机器学习、物联网和边缘计算用于健康用例的研究工作，主要是预防传染病的传播。研究的主要动机是Covid-19大流行和需要改善对预防措施实施的控制。在这项研究中，实验的重点是使用机器学习来创建和利用口罩检测的预测模型。然后在各种平台上评估预测模型，重点是在配备摄像机传感器的各种边缘设备上的使用。对标准笔记本电脑、Raspberry Pi3、Jetson Nano AI edge等不同平台进行了测试和评估。最后，本文讨论了一种可能的方法来实现一个解决方案，将利用口罩检测功能，并为未来的研究步骤奠定了路径。
{"title":"Face Mask Detection Based on Machine Learning and Edge Computing","authors":"Ivan Jovović, Dejan Babic, Stevan Cakic, Tomo Popović, S. Krco, Petar Knezevic","doi":"10.1109/INFOTEH53737.2022.9751311","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751311","url":null,"abstract":"This paper describes research effort aimed at the use of machine learning, Internet of Things, and edge computing for a use case in health, mainly the prevention of the spread of infectious diseases. The main motivation for the research was the Covid-19 pandemic and the need to improve control of the prevention measures implementation. In the study, the experimentation was focused on the use of machine learning to create and utilize prediction models for face mask detection. The prediction model is then evaluated on the various platforms with a focus on the use on various edge devices equipped with a video camera sensor. Different platforms have been tested and evaluated such as standard laptop PC, Raspberry Pi3, and Jetson Nano AI edge platform. Finally, the paper discusses a possible approach to implement a solution that would utilize the face mask detection function and lays out the path for the future research steps.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"5 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73222179","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751248
M. Maksimovic, Marijana Cosovic
The Church of the Holy Archangels Michael and Gabriel located in Sarajevo is a national monument belonging to Eastern Orthodox cultural heritage. It is a very well-preserved sacral object considering the date of first mention is 1539 and it has been used to date for the religious purposes. On the other hand, deterioration of aging historical/religious buildings is inevitable process composed of cumulative, progressive and nonlinear factors. Hence, in order to maintain their best condition for as long as possible compliance with guidelines and procedures for cultural heritage preservation is needed. Climate control within historical/religious buildings surfaced as an important research area as indoor climate is changing in recent times. Humans have always shaped their environment by desire to enjoy concurrently the comfort of modern living as well as preserve the monuments for future generations. For example, use of heating systems in historical/religious buildings are creating new challenges for their preservation. This paper is an attempt towards the implementation of Internet of Things (IoT) system with focus on preservation of the national monument using a simulation of climate control in the Church of the Holy Archangels Michael and Gabriel.
{"title":"Towards the Implementation of IoT System for Preservation: the Church of Holy Archangels Michael and Gabriel Case Study","authors":"M. Maksimovic, Marijana Cosovic","doi":"10.1109/INFOTEH53737.2022.9751248","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751248","url":null,"abstract":"The Church of the Holy Archangels Michael and Gabriel located in Sarajevo is a national monument belonging to Eastern Orthodox cultural heritage. It is a very well-preserved sacral object considering the date of first mention is 1539 and it has been used to date for the religious purposes. On the other hand, deterioration of aging historical/religious buildings is inevitable process composed of cumulative, progressive and nonlinear factors. Hence, in order to maintain their best condition for as long as possible compliance with guidelines and procedures for cultural heritage preservation is needed. Climate control within historical/religious buildings surfaced as an important research area as indoor climate is changing in recent times. Humans have always shaped their environment by desire to enjoy concurrently the comfort of modern living as well as preserve the monuments for future generations. For example, use of heating systems in historical/religious buildings are creating new challenges for their preservation. This paper is an attempt towards the implementation of Internet of Things (IoT) system with focus on preservation of the national monument using a simulation of climate control in the Church of the Holy Archangels Michael and Gabriel.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"76 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75347996","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751326
Milomir Babić, V. Risojevic
Emergence of deep learning methods during the last decade has lead to a revolution in machine learning and a significant improvement of results in various fields. Initially, these methods were based on supervised learning but, as the development progressed, the limitations stemming from the dependence on labeled datasets became apparent. Data labeling is an expensive, laborious and error prone process which is hard to automate. All this hinders the training process, especially in the applications where a large amount of data is not available. This motivated the development of different unsupervised methods that aim to utilize the wide availability of unlabeled datasets. These methods involve substitution of manual labels with data relationships which can be automatically created. In this paper we examine one such unsupervised method, contrastive multiview coding, and its application in audio classification, by adapting an implementation from the field of digital image processing. We show that the use of this method results in models which can be used for feature extraction or fine-tuned for use in different downstream tasks to achieve results that surpass the ones obtained through pure supervised learning. We also investigate the effects of domain and size of the unlabeled dataset as well as the size of the downstream dataset on the results achieved in downstream tasks through the use of frozen and fine-tuned feature extractors.
{"title":"Application of Contrastive Multiview Coding in Audio Classification","authors":"Milomir Babić, V. Risojevic","doi":"10.1109/INFOTEH53737.2022.9751326","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751326","url":null,"abstract":"Emergence of deep learning methods during the last decade has lead to a revolution in machine learning and a significant improvement of results in various fields. Initially, these methods were based on supervised learning but, as the development progressed, the limitations stemming from the dependence on labeled datasets became apparent. Data labeling is an expensive, laborious and error prone process which is hard to automate. All this hinders the training process, especially in the applications where a large amount of data is not available. This motivated the development of different unsupervised methods that aim to utilize the wide availability of unlabeled datasets. These methods involve substitution of manual labels with data relationships which can be automatically created. In this paper we examine one such unsupervised method, contrastive multiview coding, and its application in audio classification, by adapting an implementation from the field of digital image processing. We show that the use of this method results in models which can be used for feature extraction or fine-tuned for use in different downstream tasks to achieve results that surpass the ones obtained through pure supervised learning. We also investigate the effects of domain and size of the unlabeled dataset as well as the size of the downstream dataset on the results achieved in downstream tasks through the use of frozen and fine-tuned feature extractors.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"260 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77488898","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751262
M. Lazarević, G. Ostojić, D. Lukić, M. Milošević, A. Antić
Modern production systems if want to survive in the tough market must implement new technologies, which enable real-time decision making. In that way, they can react on time to overcome difficulties that arise with random probability distribution. There are different kinds of methods and technologies which are frequently used in production system processes. In this paper methods, analysis and application of different cutting-edge technologies are represented.
{"title":"Smart Production Systems: Methods and Application","authors":"M. Lazarević, G. Ostojić, D. Lukić, M. Milošević, A. Antić","doi":"10.1109/INFOTEH53737.2022.9751262","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751262","url":null,"abstract":"Modern production systems if want to survive in the tough market must implement new technologies, which enable real-time decision making. In that way, they can react on time to overcome difficulties that arise with random probability distribution. There are different kinds of methods and technologies which are frequently used in production system processes. In this paper methods, analysis and application of different cutting-edge technologies are represented.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"42 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73515196","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751308
Sara Čubrilović, Duška Mandić, Aleksandra Krstić
This paper evaluates speech-like (SL) waveform based secure voice transmission over various voice communication channels. In light of the fact that different voice channels exhibit different problems during transmission in terms of non-linearity, signal compression, etc. and call quality is heavily dependent on the type of communication equipment, conventional SL waveform based codebook (CB) as a unique solution was deemed insufficient for reliable secure communication. Principal improvement is in introduction of more sophisticated symbol classification and recognition technique that led to a significant error reduction, rendering this secure transmission method applicable in multiple voice communication scenarios. Simulations that were carried out in real-time scenarios with genuine equipment confirmed the supremacy of the proposed modification over the conventional solution.
{"title":"Evaluation of Improved Classification of Speech-Like Waveforms Used for Secure Voice Transmission","authors":"Sara Čubrilović, Duška Mandić, Aleksandra Krstić","doi":"10.1109/INFOTEH53737.2022.9751308","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751308","url":null,"abstract":"This paper evaluates speech-like (SL) waveform based secure voice transmission over various voice communication channels. In light of the fact that different voice channels exhibit different problems during transmission in terms of non-linearity, signal compression, etc. and call quality is heavily dependent on the type of communication equipment, conventional SL waveform based codebook (CB) as a unique solution was deemed insufficient for reliable secure communication. Principal improvement is in introduction of more sophisticated symbol classification and recognition technique that led to a significant error reduction, rendering this secure transmission method applicable in multiple voice communication scenarios. Simulations that were carried out in real-time scenarios with genuine equipment confirmed the supremacy of the proposed modification over the conventional solution.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"1 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"75664690","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751253
V. Kuchanskyy, O. Rubanenko, Marijana Cosovic, I. Hunko
The possibilities of using artificial neural networks (ANNs) for quick decision-making in the events of prolonged surges are presented in this paper considering that neural networks can establish non-linear relationships between the parameters of an ultra-high voltage transmission line. Research has been carried out based on theoretical models as well as practical problems aiming at the analysis of resonant overvoltages during their occurrence, development and existence. Determining of overvoltage characteristics was carried out in the presence of a significant number of fuzzy specified factors affecting the accuracy. The multilayer model, suitable for identifying the factors having the greatest impact on the occurrence, frequency and multiplicity of overvoltages in electrical networks, is applied. The resonant overvoltages were generated by connecting the autotransformer to the electrical bulk network. The results of determining the characteristics of resonant overvoltages using ANNs are presented in this paper. To achieve this goal, the following four tasks were formulated: (i) overvoltage characteristics using neural network methods were determined, (ii) neural network model corresponding to power line initial data was built, (iii) forecasted results were obtained, and (iv) the accuracy of constructed model was evaluated.
{"title":"Analyzing the Effects of Abnormal Resonance Voltages using Artificial Neural Networks","authors":"V. Kuchanskyy, O. Rubanenko, Marijana Cosovic, I. Hunko","doi":"10.1109/INFOTEH53737.2022.9751253","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751253","url":null,"abstract":"The possibilities of using artificial neural networks (ANNs) for quick decision-making in the events of prolonged surges are presented in this paper considering that neural networks can establish non-linear relationships between the parameters of an ultra-high voltage transmission line. Research has been carried out based on theoretical models as well as practical problems aiming at the analysis of resonant overvoltages during their occurrence, development and existence. Determining of overvoltage characteristics was carried out in the presence of a significant number of fuzzy specified factors affecting the accuracy. The multilayer model, suitable for identifying the factors having the greatest impact on the occurrence, frequency and multiplicity of overvoltages in electrical networks, is applied. The resonant overvoltages were generated by connecting the autotransformer to the electrical bulk network. The results of determining the characteristics of resonant overvoltages using ANNs are presented in this paper. To achieve this goal, the following four tasks were formulated: (i) overvoltage characteristics using neural network methods were determined, (ii) neural network model corresponding to power line initial data was built, (iii) forecasted results were obtained, and (iv) the accuracy of constructed model was evaluated.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"45 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"73748372","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751283
I. Grobelna, A. Karatkevich
Petri nets are a powerful technique for modelling flexible manufacturing systems. However, in some situations the system may get stuck in a deadlock state and suspend its operation mode. Here, we propose a novel deadlock recovery policy that may be used to automatically recover from the deadlock states, based on the analysis of a full reachability graph with minimized traversing. Additional recovery transitions are added to the existing structure of a Petri net without changing the existing state space. The solution may not be optimal regarding the number of added recovery transitions, but it can be found in a simple way by considering the closest legal markings. In the paper, the newly proposed method is also illustrated with a case study and compared to the other existing approaches. The preliminary results show, that despite its simplicity, the found deadlock recovery solution is comparable to other more complex methods from the literature, regarding the number of added recovery transitions.
{"title":"A Deadlock Recovery Policy for Flexible Manufacturing Systems with Minimized Traversing within Reachability Graph","authors":"I. Grobelna, A. Karatkevich","doi":"10.1109/INFOTEH53737.2022.9751283","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751283","url":null,"abstract":"Petri nets are a powerful technique for modelling flexible manufacturing systems. However, in some situations the system may get stuck in a deadlock state and suspend its operation mode. Here, we propose a novel deadlock recovery policy that may be used to automatically recover from the deadlock states, based on the analysis of a full reachability graph with minimized traversing. Additional recovery transitions are added to the existing structure of a Petri net without changing the existing state space. The solution may not be optimal regarding the number of added recovery transitions, but it can be found in a simple way by considering the closest legal markings. In the paper, the newly proposed method is also illustrated with a case study and compared to the other existing approaches. The preliminary results show, that despite its simplicity, the found deadlock recovery solution is comparable to other more complex methods from the literature, regarding the number of added recovery transitions.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"7 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74368929","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751317
Nikolas Naydenov, Stela Ruseva
Containerization is a virtualization technology that facilitates the deployment of applications. Container Orchestration is the process of automating the deployment, management, scaling and networking of containers. In this systematic mapping study, we are presenting the analysis of recent scientific papers that deal with containerization and container orchestration in the cloud, combined with machine learning, and how these are utilized to solve problems in different application areas. Currently new challenges arise related to the processing of big data, but also the optimized management of increasing amount of heterogeneous workloads in a cloud environment. The analysis results from the publications of recent years show the growing interest in the scientific community in these evolving technologies - container orchestration from one hand and utilizing machine learning on the other. The emphasis of the study are the trends and innovations, the orchestration technologies and strategies, the machine learning algorithms. Evaluating the relevance of the proposed solutions and ideas for future research are also outlined.
{"title":"Combining Container Orchestration and Machine Learning in the Cloud: a Systematic Mapping Study","authors":"Nikolas Naydenov, Stela Ruseva","doi":"10.1109/INFOTEH53737.2022.9751317","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751317","url":null,"abstract":"Containerization is a virtualization technology that facilitates the deployment of applications. Container Orchestration is the process of automating the deployment, management, scaling and networking of containers. In this systematic mapping study, we are presenting the analysis of recent scientific papers that deal with containerization and container orchestration in the cloud, combined with machine learning, and how these are utilized to solve problems in different application areas. Currently new challenges arise related to the processing of big data, but also the optimized management of increasing amount of heterogeneous workloads in a cloud environment. The analysis results from the publications of recent years show the growing interest in the scientific community in these evolving technologies - container orchestration from one hand and utilizing machine learning on the other. The emphasis of the study are the trends and innovations, the orchestration technologies and strategies, the machine learning algorithms. Evaluating the relevance of the proposed solutions and ideas for future research are also outlined.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"70 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"77764877","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751294
Tamara Skoric, D. Bajić
The two-photon laser scanning microscopic (TPLSM) image does not have an adequate quality test because it lacks a real ground truth image for comparing with the denoised image. The paper proposes an artificial test image to test the quality of denoising techniques. The image is generated as a modified Chessboard with simulated mixed Poison-Gauss noise, in which there are fields with three different shades. The possibilities of twelve state-of-the-art methods for noise reduction on the proposed quality test image and TPLSM image in publicly available databases were tested. Many methods are not able to distinguish between different fields of a Chessboard test image with a low signal-to-noise ratio. The instability of methods, that were not originally developed for the reduction of the Poison-Gaussian noise, was also confirmed in the proposed test image as well as the TPLSM images.
{"title":"Noise reduction quality test for two-photon laser scanning microscopic images","authors":"Tamara Skoric, D. Bajić","doi":"10.1109/INFOTEH53737.2022.9751294","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751294","url":null,"abstract":"The two-photon laser scanning microscopic (TPLSM) image does not have an adequate quality test because it lacks a real ground truth image for comparing with the denoised image. The paper proposes an artificial test image to test the quality of denoising techniques. The image is generated as a modified Chessboard with simulated mixed Poison-Gauss noise, in which there are fields with three different shades. The possibilities of twelve state-of-the-art methods for noise reduction on the proposed quality test image and TPLSM image in publicly available databases were tested. Many methods are not able to distinguish between different fields of a Chessboard test image with a low signal-to-noise ratio. The instability of methods, that were not originally developed for the reduction of the Poison-Gaussian noise, was also confirmed in the proposed test image as well as the TPLSM images.","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"25 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"82152496","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 : 2022-03-16DOI: 10.1109/INFOTEH53737.2022.9751279
Jakob Triva, R. Grbić, M. Vranješ, N. Teslic
The current environmental conditions should be monitored during autonomous driving since the different weather conditions can have a different impact on implemented sensor system or on the efficiency of the implemented control system. In this paper, the classification of weather conditions in the vehicle environment is based on images captured by a front-view camera, which are further processed by the simple Convolutional Neural Network (CNN). For model development purposes, training and validation data sets were created from two sources: the BDD100K database and by extracting frames from the collected video sequences. The solution implements an additional mechanism to filter out false predictions based on a circular buffer. The proposed solution achieves the F1 measure of 98.3% for the entire test video frames data set, where it achieves the best results in snowy weather detection (Precision of 100%, F1 of 100.00%) and the worst in foggy weather detection (Precision of 97.25%, F1 of 98.00%).
{"title":"Weather Condition Classification in Vehicle Environment Based on Front-View Camera Images","authors":"Jakob Triva, R. Grbić, M. Vranješ, N. Teslic","doi":"10.1109/INFOTEH53737.2022.9751279","DOIUrl":"https://doi.org/10.1109/INFOTEH53737.2022.9751279","url":null,"abstract":"The current environmental conditions should be monitored during autonomous driving since the different weather conditions can have a different impact on implemented sensor system or on the efficiency of the implemented control system. In this paper, the classification of weather conditions in the vehicle environment is based on images captured by a front-view camera, which are further processed by the simple Convolutional Neural Network (CNN). For model development purposes, training and validation data sets were created from two sources: the BDD100K database and by extracting frames from the collected video sequences. The solution implements an additional mechanism to filter out false predictions based on a circular buffer. The proposed solution achieves the F1 measure of 98.3% for the entire test video frames data set, where it achieves the best results in snowy weather detection (Precision of 100%, F1 of 100.00%) and the worst in foggy weather detection (Precision of 97.25%, F1 of 98.00%).","PeriodicalId":6839,"journal":{"name":"2022 21st International Symposium INFOTEH-JAHORINA (INFOTEH)","volume":"37 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2022-03-16","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"84284152","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：