Pub Date : 2022-08-31DOI: 10.1109/CSI54720.2022.9924131
S. T. S., P. Sreeja, Rajeev J Ram
Because of the exponential expansion of knowledge available on the internet, it is becoming impossible to decipher Real News from false News. Thus, this contributes to the spread of false information. Many dangerous fake accounts have been created recently, and these accounts distribute false information via posts, blogs, etc. across social media. Some people spread this false information without being aware of its falsity. In this proposal, we proposed a model to identify the fake news spreading on social media. To accomplish this model, we collected the dataset named “NEWS” from the Kaggle depository. Machine learning algorithms such as Random Forest, Passive Aggressive, and Gradient Boosting were used to Classify Real News and Fake News from News Articles. The passive Aggressive Algorithm gave better accuracy than the other two Algorithms used in this work.
{"title":"Fake News Article classification using Random Forest, Passive Aggressive, and Gradient Boosting","authors":"S. T. S., P. Sreeja, Rajeev J Ram","doi":"10.1109/CSI54720.2022.9924131","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924131","url":null,"abstract":"Because of the exponential expansion of knowledge available on the internet, it is becoming impossible to decipher Real News from false News. Thus, this contributes to the spread of false information. Many dangerous fake accounts have been created recently, and these accounts distribute false information via posts, blogs, etc. across social media. Some people spread this false information without being aware of its falsity. In this proposal, we proposed a model to identify the fake news spreading on social media. To accomplish this model, we collected the dataset named “NEWS” from the Kaggle depository. Machine learning algorithms such as Random Forest, Passive Aggressive, and Gradient Boosting were used to Classify Real News and Fake News from News Articles. The passive Aggressive Algorithm gave better accuracy than the other two Algorithms used in this work.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"11 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"115412137","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924024
Tiya Ann Siby, Sonam Pal, Jessica Arlina, S. Nagaraju
Real-Time Sign Language Recognition (RTSLG) can help people express clearer thoughts, speak in shorter sentences, and be more expressive to use declarative language. Hand gestures provide a wealth of information that persons with disabilities can use to communicate in a fundamental way and to complement communication for others. Since the hand gesture information is based on movement sequences, accurately detecting hand gestures in real-time is difficult. Hearing-impaired persons have difficulty interacting with others, resulting in a communication gap. The only way for them to communicate their ideas and feelings is to use hand signals, which are not understood by many people. As a result, in recent days, the hand gesture detection system has gained prominence. In this paper, the proposed design is of a deep learning model using Python, TensorFlow, OpenCV and Histogram Equalization that can be accessed from the web browser. The proposed RTSLG system uses image detection, computer vision, and neural network methodologies i.e. Convolution Neural Network to recognise the characteristics of the hand in video filmed by a web camera. To enhance the details of the images, an image processing technique called Histogram Equalization is performed. The accuracy obtained by the proposed system is 87.8%. Once the gesture is recognized and text output is displayed, the proposed RTSLG system makes use of gTTS (Google Text-to-Speech) library in order to convert the displayed text to audio for assisting the communication of speech and hearing-impaired person.
实时手语识别(RTSLG)可以帮助人们表达更清晰的思想，用更短的句子说话，并且更善于使用陈述性语言。手势提供了丰富的信息，残疾人可以利用这些信息进行基本的交流，并补充他人的交流。由于手势信息是基于动作序列的，因此很难实时准确地检测手势。听力受损的人与他人交流有困难，导致沟通障碍。他们表达想法和感受的唯一方式就是用手势，而很多人都不懂手势。因此，最近几天，手势检测系统得到了重视。在本文中，提出的设计是一个使用Python, TensorFlow, OpenCV和直方图均衡化的深度学习模型，可以从web浏览器访问。提出的RTSLG系统使用图像检测、计算机视觉和神经网络方法，即卷积神经网络来识别由网络摄像机拍摄的视频中的手的特征。为了增强图像的细节，执行了一种称为直方图均衡化的图像处理技术。该系统的精度为87.8%。一旦手势被识别并显示文本输出，所提出的RTSLG系统利用gTTS (Google text -to- speech)库将显示的文本转换为音频，以帮助言语和听力受损人士进行交流。
{"title":"Gesture based Real-Time Sign Language Recognition System","authors":"Tiya Ann Siby, Sonam Pal, Jessica Arlina, S. Nagaraju","doi":"10.1109/CSI54720.2022.9924024","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924024","url":null,"abstract":"Real-Time Sign Language Recognition (RTSLG) can help people express clearer thoughts, speak in shorter sentences, and be more expressive to use declarative language. Hand gestures provide a wealth of information that persons with disabilities can use to communicate in a fundamental way and to complement communication for others. Since the hand gesture information is based on movement sequences, accurately detecting hand gestures in real-time is difficult. Hearing-impaired persons have difficulty interacting with others, resulting in a communication gap. The only way for them to communicate their ideas and feelings is to use hand signals, which are not understood by many people. As a result, in recent days, the hand gesture detection system has gained prominence. In this paper, the proposed design is of a deep learning model using Python, TensorFlow, OpenCV and Histogram Equalization that can be accessed from the web browser. The proposed RTSLG system uses image detection, computer vision, and neural network methodologies i.e. Convolution Neural Network to recognise the characteristics of the hand in video filmed by a web camera. To enhance the details of the images, an image processing technique called Histogram Equalization is performed. The accuracy obtained by the proposed system is 87.8%. Once the gesture is recognized and text output is displayed, the proposed RTSLG system makes use of gTTS (Google Text-to-Speech) library in order to convert the displayed text to audio for assisting the communication of speech and hearing-impaired person.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"15 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129014256","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924143
U. K. Agrawal, A. Shrivastava, Debanjan Das, R. Mahapatra
In fast varying channels for Mobile molecular communication (MMC), detection is not easy. This challenge exists due to quick changes in the channel impulse response (CIR) in diffusive environment. The conventional detectors require channel state information (CSI) for accurate detection in MMC. Since, CSI is difficult to obtain in fast varying channels, the present research work proposes a neural network detector (NND) that does not require CSI in MMC, even for the channels varying rapidly. The NND uses BFGS algorithm for optimizing its weights. The performance of NND is determined using the data driven approach for training and testing. The bit error rate (BER) has been found for different numbers of nodes and layers. The optimized approach is carried out to trade off between computational burden and BER by variation of nodes as well as layers of the NND. In case of signal to noise ratio (SNR) of 39 dB, our network performs better than existing works in the literature.
{"title":"Neural Network Detector in Mobile Molecular Communication for Fast Varying Channels","authors":"U. K. Agrawal, A. Shrivastava, Debanjan Das, R. Mahapatra","doi":"10.1109/CSI54720.2022.9924143","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924143","url":null,"abstract":"In fast varying channels for Mobile molecular communication (MMC), detection is not easy. This challenge exists due to quick changes in the channel impulse response (CIR) in diffusive environment. The conventional detectors require channel state information (CSI) for accurate detection in MMC. Since, CSI is difficult to obtain in fast varying channels, the present research work proposes a neural network detector (NND) that does not require CSI in MMC, even for the channels varying rapidly. The NND uses BFGS algorithm for optimizing its weights. The performance of NND is determined using the data driven approach for training and testing. The bit error rate (BER) has been found for different numbers of nodes and layers. The optimized approach is carried out to trade off between computational burden and BER by variation of nodes as well as layers of the NND. In case of signal to noise ratio (SNR) of 39 dB, our network performs better than existing works in the literature.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124249704","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924018
Samprit Bose, Chavan Deep Ramesh, M. Kolekar
Traffic has been a major concern in most of the cities. Monitoring cameras are used to track, detect and count vehicles in real-time to ensure proper management of traffic. Counting of vehicles like cars, trucks and two wheelers is important for Intelligent Transportation System (ITS) to identify the intensity of traffic flow. In this paper we proposed vision based vehicle classification and counting approach using YOLO- v3 framework. The proposed method is composed of steps like masking, detection, classification and counting of different classes of vehicles. We have tested proposed method over 2000 vehicles of different categories obtained from the CCTV camera installed at main gate of lIT Patna campus. Experimental results show that the proposed approach has achieved accuracy of 93.65 % and 87.68 % during daytime and nighttime respectively.
{"title":"Vehicle Classification and Counting for Traffic Video Monitoring Using YOLO-v3","authors":"Samprit Bose, Chavan Deep Ramesh, M. Kolekar","doi":"10.1109/CSI54720.2022.9924018","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924018","url":null,"abstract":"Traffic has been a major concern in most of the cities. Monitoring cameras are used to track, detect and count vehicles in real-time to ensure proper management of traffic. Counting of vehicles like cars, trucks and two wheelers is important for Intelligent Transportation System (ITS) to identify the intensity of traffic flow. In this paper we proposed vision based vehicle classification and counting approach using YOLO- v3 framework. The proposed method is composed of steps like masking, detection, classification and counting of different classes of vehicles. We have tested proposed method over 2000 vehicles of different categories obtained from the CCTV camera installed at main gate of lIT Patna campus. Experimental results show that the proposed approach has achieved accuracy of 93.65 % and 87.68 % during daytime and nighttime respectively.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"96 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124532334","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924098
Ashwini Sasi, Varun Nair, Vipin P
The darkweb has always been a supreme source for conducting web-based illegal activities. Cyber-criminals have been consistently using the darkweb for providing malicious services and carrying out criminal activities. With a multitude of networks that are volunteered, it is hard to trace down these malicious services and service providers. OSNIT is a method that is used to accumulate information on a specific targeted entity by gathering the data available publicly. Implementing open source intelligence on the Tor-based network is a difficult task for both developers and researchers due to the sophisticated onion routing feature. Onion routing provides a strong anonymity feature for its users. A tool is introduced through this work for OSNIT on the darkweb. The tool is expected to help law enforcement agencies and threat researchers to automate the task of extraction of pictorial intelligence from different malicious sites in the darkweb.
{"title":"DARKWEB IMAGE SCRAPPER: An Open Source Intelligence Tool","authors":"Ashwini Sasi, Varun Nair, Vipin P","doi":"10.1109/CSI54720.2022.9924098","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924098","url":null,"abstract":"The darkweb has always been a supreme source for conducting web-based illegal activities. Cyber-criminals have been consistently using the darkweb for providing malicious services and carrying out criminal activities. With a multitude of networks that are volunteered, it is hard to trace down these malicious services and service providers. OSNIT is a method that is used to accumulate information on a specific targeted entity by gathering the data available publicly. Implementing open source intelligence on the Tor-based network is a difficult task for both developers and researchers due to the sophisticated onion routing feature. Onion routing provides a strong anonymity feature for its users. A tool is introduced through this work for OSNIT on the darkweb. The tool is expected to help law enforcement agencies and threat researchers to automate the task of extraction of pictorial intelligence from different malicious sites in the darkweb.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"32 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"134147361","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924019
Zitong Li, T. Lin, Xia Zhao
Buy and sell signals in stock trends are related to the yield of investors. In this paper, trading signal filtering is regarded as a binary classification problem, and a stock trading signal filtering model based on Bagging, Random Forest(RF) and Logistic Regression(LR) is proposed. Firstly, the trading information of different indexes in the stock market is mined according to the selected attributes. Secondly, the optimal number of features is selected according to the comparative experimental results. Finally, a multi-classifier ensemble model is built, which based on Bagaging-RF -LR. The trading signals are put into the model, and the soft voting method is used to learn and classify the data. The experimental results show that the classification accuracy of the ensemble model reaches 61%, which is 1 % $sim$ 2 % higher than that of the single classification model, and the mean ration increases from 145.19 % $sim$ 166.48 % to 171.01%. The comparison of the experimental results shows that the Bagaging-RF-LR model is effective and has a good classification effect on the trading signal filtering problem.
{"title":"Stock Trading Signal Filtering Based on Bagging-RF-LR Model","authors":"Zitong Li, T. Lin, Xia Zhao","doi":"10.1109/CSI54720.2022.9924019","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924019","url":null,"abstract":"Buy and sell signals in stock trends are related to the yield of investors. In this paper, trading signal filtering is regarded as a binary classification problem, and a stock trading signal filtering model based on Bagging, Random Forest(RF) and Logistic Regression(LR) is proposed. Firstly, the trading information of different indexes in the stock market is mined according to the selected attributes. Secondly, the optimal number of features is selected according to the comparative experimental results. Finally, a multi-classifier ensemble model is built, which based on Bagaging-RF -LR. The trading signals are put into the model, and the soft voting method is used to learn and classify the data. The experimental results show that the classification accuracy of the ensemble model reaches 61%, which is 1 % $sim$ 2 % higher than that of the single classification model, and the mean ration increases from 145.19 % $sim$ 166.48 % to 171.01%. The comparison of the experimental results shows that the Bagaging-RF-LR model is effective and has a good classification effect on the trading signal filtering problem.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"4 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133739081","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924103
Raushan Kumar Singh, Akshay Kumar, M. Hussain
In the modern era of society there is a growing usage of the Internet of Things and its applications can be seen in our daily lives like Smart home, life has become easy and comfortable but there is a tradeoff between security and privacy. Hence there is a need for efficient and secure security mechanisms for Smart homes to further elevate their usage. We have proposed a key agreement scheme between the devices in a smart home. The proposed mechanism established a symmetric key between the base station and the device securely. The established key can be further used for the secure transmission of data. The proposed mechanism is verified for security using the Scyther tool and found that it is secure. As we are not using any complex operation to generate a symmetric key the proposed scheme is lightweight.
{"title":"A Secure Key Agreement Mechanism for Smart Home Networks","authors":"Raushan Kumar Singh, Akshay Kumar, M. Hussain","doi":"10.1109/CSI54720.2022.9924103","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924103","url":null,"abstract":"In the modern era of society there is a growing usage of the Internet of Things and its applications can be seen in our daily lives like Smart home, life has become easy and comfortable but there is a tradeoff between security and privacy. Hence there is a need for efficient and secure security mechanisms for Smart homes to further elevate their usage. We have proposed a key agreement scheme between the devices in a smart home. The proposed mechanism established a symmetric key between the base station and the device securely. The established key can be further used for the secure transmission of data. The proposed mechanism is verified for security using the Scyther tool and found that it is secure. As we are not using any complex operation to generate a symmetric key the proposed scheme is lightweight.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"45 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124684791","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9924000
N. Behera, Monoranjan Pradhan, P. Mishro
In a Very Large Scale Integration (VLSI) Field, multipliers play a vital role. In practice, multipliers are utilizing as unsigned and signed category. An unsigned multiplier does the multiplication of two unsigned binary integers, whereas in signed multiplier the multiplication is done by every bit of binary integers. It can be expanded within its series or a suitable outcome. In the prose, most of the research has been stated that describes signed multiplication such as Booth, Baugh-Wooley, Wallace tree, Array multiplier proposed the elevated speed signed product procedures. However, there is a possibility will get the better performance of delay, power and speed in the signed multiplier. In this paper, author proposed a signed multiplier utilizing “Urdhva Tiryabhyam” (UT) algorithm. The suggested intend structure is appropriate for the conversion of signed and unsigned binary multiplication and decimal multiplication. Using of Vedic algorithm in the suggested work, the system performance is progressed and the area is minimized. The proposed structure is simulated and synthesized using ISE Xilinx 14.5 and implemented in Virtex 4 Field Programmable Gate Array devices (FPGA). The recommended work is compared among the prior architectures. From the outcomes the greatest of the author's design is taken.
{"title":"Analysis of Combinational Delay in Signed Binary Multiplier","authors":"N. Behera, Monoranjan Pradhan, P. Mishro","doi":"10.1109/CSI54720.2022.9924000","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9924000","url":null,"abstract":"In a Very Large Scale Integration (VLSI) Field, multipliers play a vital role. In practice, multipliers are utilizing as unsigned and signed category. An unsigned multiplier does the multiplication of two unsigned binary integers, whereas in signed multiplier the multiplication is done by every bit of binary integers. It can be expanded within its series or a suitable outcome. In the prose, most of the research has been stated that describes signed multiplication such as Booth, Baugh-Wooley, Wallace tree, Array multiplier proposed the elevated speed signed product procedures. However, there is a possibility will get the better performance of delay, power and speed in the signed multiplier. In this paper, author proposed a signed multiplier utilizing “Urdhva Tiryabhyam” (UT) algorithm. The suggested intend structure is appropriate for the conversion of signed and unsigned binary multiplication and decimal multiplication. Using of Vedic algorithm in the suggested work, the system performance is progressed and the area is minimized. The proposed structure is simulated and synthesized using ISE Xilinx 14.5 and implemented in Virtex 4 Field Programmable Gate Array devices (FPGA). The recommended work is compared among the prior architectures. From the outcomes the greatest of the author's design is taken.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"8 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130953632","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9923993
M. Mathews, S. M. Anzar
Diabetic macular edema (DME) and drusen macular degeneration (DMD) are two se-vere vision-threatening diseases affecting the mac-ula of the eye. This study presents deep learning based classification models for retinal optical co-herence tomography (OCT) images that distinguish between healthy eyes, DME and DMD cases. The work involves the use of residual models, and deep and densely connected networks for the analysis of OCT images. This involves pre-initialisation of the model using the ImageNet dataset, followed by fine-tuning with OCT images. Both models are very powerful and suitable for real-time use in clinical practise. ResidualNets use skip connections, where the output of the previous layer is added to the layer before it. DenseNets use dense connections between the convolutional layers of the network, which allows deeper supervision between layers. This makes it easier for the model to learn the complex feature maps of the images of OCT in each layer of the network. The models are trained and evaluated using the Mendeley OCT dataset, a publicly available SD-OCT dataset for the retina. We calculate the F1 score, accuracy, precision and recall to evaluate the models. The models provide excellent performance without requiring any pre-processing steps. The promising performance of the computerised systems prove that they can serve as automatic recognition tools to assist ophthalmologists.
{"title":"Residual Networks and Deep-Densely Connected Networks for the Classification of retinal OCT Images","authors":"M. Mathews, S. M. Anzar","doi":"10.1109/CSI54720.2022.9923993","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9923993","url":null,"abstract":"Diabetic macular edema (DME) and drusen macular degeneration (DMD) are two se-vere vision-threatening diseases affecting the mac-ula of the eye. This study presents deep learning based classification models for retinal optical co-herence tomography (OCT) images that distinguish between healthy eyes, DME and DMD cases. The work involves the use of residual models, and deep and densely connected networks for the analysis of OCT images. This involves pre-initialisation of the model using the ImageNet dataset, followed by fine-tuning with OCT images. Both models are very powerful and suitable for real-time use in clinical practise. ResidualNets use skip connections, where the output of the previous layer is added to the layer before it. DenseNets use dense connections between the convolutional layers of the network, which allows deeper supervision between layers. This makes it easier for the model to learn the complex feature maps of the images of OCT in each layer of the network. The models are trained and evaluated using the Mendeley OCT dataset, a publicly available SD-OCT dataset for the retina. We calculate the F1 score, accuracy, precision and recall to evaluate the models. The models provide excellent performance without requiring any pre-processing steps. The promising performance of the computerised systems prove that they can serve as automatic recognition tools to assist ophthalmologists.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121528236","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 : 2022-08-31DOI: 10.1109/CSI54720.2022.9923977
A. Sreejithlal, M. N. Syam, T. M. Letha, K. P. M. Madhusoodanan, R. Warrier, A. Shooja
This paper discusses the design of an automated, low-cost and portable system intended to test multiple heterogenous sensors in a system. Systems deals mainly with providing excitation to the pressure and temperature sensors, the signal conditioning of sensor output and displaying the results on an operator-interactive screen. It combines capability to carry out multiple tests on the sensors for determining sensor health; as chosen by the operator. Results of the tests are compared with the expected values and are stored in memory for retrieval. Design is based on Raspberry pi which presents a user-interface to the operator. A set of Arduino-pro mini boards perform the sensor line selection by switching a relay network and carry out source selection for providing sensor excitation.
{"title":"Automated sensor test system using Raspberry Pi","authors":"A. Sreejithlal, M. N. Syam, T. M. Letha, K. P. M. Madhusoodanan, R. Warrier, A. Shooja","doi":"10.1109/CSI54720.2022.9923977","DOIUrl":"https://doi.org/10.1109/CSI54720.2022.9923977","url":null,"abstract":"This paper discusses the design of an automated, low-cost and portable system intended to test multiple heterogenous sensors in a system. Systems deals mainly with providing excitation to the pressure and temperature sensors, the signal conditioning of sensor output and displaying the results on an operator-interactive screen. It combines capability to carry out multiple tests on the sensors for determining sensor health; as chosen by the operator. Results of the tests are compared with the expected values and are stored in memory for retrieval. Design is based on Raspberry pi which presents a user-interface to the operator. A set of Arduino-pro mini boards perform the sensor line selection by switching a relay network and carry out source selection for providing sensor excitation.","PeriodicalId":221137,"journal":{"name":"2022 International Conference on Connected Systems & Intelligence (CSI)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-08-31","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"133929470","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：