Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563328
P. S. Hitha, G. Ragesh, Dr. Anish R
Image compression is a fundamental technique in digital image processing used to decrease the space used for storage of digital images and videos, which will help to increase the storage space and for efficient transmission. Nowadays many deep learning techniques have produced promising results on image compression field. However, traditional compression techniques have introduced many compression artifacts problem. To solve this problem we have compared two deep learning approaches for image compression. One method is based on Deep Autoencoder technique and other is based on deep convolutional neural network (deep CNN) approach. Autoencoder structure is a popular choice to do end-to-end compression and deep CNN is the most popular neural network model for the application of any basic deep learning technique. The performance of two methods are compared based on Peak signal-to-noise ratio (PSNR) and root mean square error (RMSE). Based on the performance evaluation methods result, it is evident that deep Autoencoder technique is more advantageous than deep CNN technique.
{"title":"Comparison Of Image Compression Analysis Using Deep Autoencoder And Deep CNN Approach","authors":"P. S. Hitha, G. Ragesh, Dr. Anish R","doi":"10.1109/ACCESS51619.2021.9563328","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563328","url":null,"abstract":"Image compression is a fundamental technique in digital image processing used to decrease the space used for storage of digital images and videos, which will help to increase the storage space and for efficient transmission. Nowadays many deep learning techniques have produced promising results on image compression field. However, traditional compression techniques have introduced many compression artifacts problem. To solve this problem we have compared two deep learning approaches for image compression. One method is based on Deep Autoencoder technique and other is based on deep convolutional neural network (deep CNN) approach. Autoencoder structure is a popular choice to do end-to-end compression and deep CNN is the most popular neural network model for the application of any basic deep learning technique. The performance of two methods are compared based on Peak signal-to-noise ratio (PSNR) and root mean square error (RMSE). Based on the performance evaluation methods result, it is evident that deep Autoencoder technique is more advantageous than deep CNN technique.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"26 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126589953","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563298
Bosco Paul Alapatt, Felix M. Philip, Anupama Jims
In recent times, fiber optic communication networks have become commonly applied for commercial as well as military applications. Fiber optic networks have gained popularity owing to the high data rate. At the same time, the generation of huge quantity of data at a faster rate poses a major challenge in the storing and transmission process. To resolve this issue, data compression approaches have been presented to reduce the quantity of transmitted data and thereby minimizes bandwidth utilization and memory. Vector quantization (VQ) is a commonly employed image compression technique and Linde Buzo Gray (LBG) is used to construct an optimum codebook to compress images. With this motivation, this paper presents a new oppositional glowworm swarm optimization based LBG (OGSO-LBG) technique for image compression in fiber optic communication. The OGSO algorithm involves the integration of oppositional based learning (OBL) concept into the GSO algorithm to boost its convergence rate. The OGSO-LBG algorithm produces the codebook at a faster rate with minimal computation complexity. In order to highlight the enhanced compression performance of the OGSO-LBG technique, a series of experiments were carried out and the results are examined under different dimensions.
{"title":"Oppositional Glowworm Swarm based Vector Quantization Technique for Image Compression in Fiber Optic Communication","authors":"Bosco Paul Alapatt, Felix M. Philip, Anupama Jims","doi":"10.1109/ACCESS51619.2021.9563298","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563298","url":null,"abstract":"In recent times, fiber optic communication networks have become commonly applied for commercial as well as military applications. Fiber optic networks have gained popularity owing to the high data rate. At the same time, the generation of huge quantity of data at a faster rate poses a major challenge in the storing and transmission process. To resolve this issue, data compression approaches have been presented to reduce the quantity of transmitted data and thereby minimizes bandwidth utilization and memory. Vector quantization (VQ) is a commonly employed image compression technique and Linde Buzo Gray (LBG) is used to construct an optimum codebook to compress images. With this motivation, this paper presents a new oppositional glowworm swarm optimization based LBG (OGSO-LBG) technique for image compression in fiber optic communication. The OGSO algorithm involves the integration of oppositional based learning (OBL) concept into the GSO algorithm to boost its convergence rate. The OGSO-LBG algorithm produces the codebook at a faster rate with minimal computation complexity. In order to highlight the enhanced compression performance of the OGSO-LBG technique, a series of experiments were carried out and the results are examined under different dimensions.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"38 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"127062964","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563300
P. Athira, N. Deepa, Nimmy M Philip, E. G. Anoop
Modular adders are used in various applications of computer systems. Modular addition is commonly used in residue number system processors. It is used mainly in the residue arithmetic unit and also in both the forward and reverse converters. Residue number system is highly efficient when compared with positional number system, because it provides high speed computation as well as less area requirement. In order to improve the computation speed, efficient modular adders are required. Modular adders based on thermometer code residue and one hot code residue are used for this purpose. This results in less latency and area. This approach reduces the area and delay of modular adders since there is no carry bit propagation during modular addition operation. It also simplifies the structure of modular adders compared to conventional binary based modular adders. All the proposed modular adders are described in verilog HDL and verified using Xilinx ISE.
{"title":"Modular Adder Designs Based On Thermometer Coding And One-Hot Coding","authors":"P. Athira, N. Deepa, Nimmy M Philip, E. G. Anoop","doi":"10.1109/ACCESS51619.2021.9563300","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563300","url":null,"abstract":"Modular adders are used in various applications of computer systems. Modular addition is commonly used in residue number system processors. It is used mainly in the residue arithmetic unit and also in both the forward and reverse converters. Residue number system is highly efficient when compared with positional number system, because it provides high speed computation as well as less area requirement. In order to improve the computation speed, efficient modular adders are required. Modular adders based on thermometer code residue and one hot code residue are used for this purpose. This results in less latency and area. This approach reduces the area and delay of modular adders since there is no carry bit propagation during modular addition operation. It also simplifies the structure of modular adders compared to conventional binary based modular adders. All the proposed modular adders are described in verilog HDL and verified using Xilinx ISE.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"216 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132049746","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563341
Joshma Joby, Robin Jose Raju, Roshan Job, Sandra Merin Thomas, A. S*
In recent years, several models based on deep learning have been proposed for the identification of Pneumonia from X-ray image of lungs. Lack of datasets with appropriate number of training images is the major challenge faced by these automated models. In this paper, we propose a model called PneumoGAN that not only augments the training dataset by generating enough number of chest X-ray images from random noise but also has the ability to detect pneumonia from a previously unseen image. The proposed model is inspired from Generative Adversarial Networks (GANs). The discriminator of the proposed PneumoGAN model involves five layers while the generator has six layers in it. The experimental results demonstrate the fact that PneumoGAN has precision, recall and F1 score of 87.71%, 91.4% and 89.52% respectively on benchmark datasets. Moreover, an AUC value of 85% is yielded by the proposed approach. Hence, the proposed model helps doctors to speed up the diagnosis process and narrowing the time required to determine whether a person is a pneumonia victim.
{"title":"PneumoGAN: A GAN based Model for Pneumonia Detection","authors":"Joshma Joby, Robin Jose Raju, Roshan Job, Sandra Merin Thomas, A. S*","doi":"10.1109/ACCESS51619.2021.9563341","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563341","url":null,"abstract":"In recent years, several models based on deep learning have been proposed for the identification of Pneumonia from X-ray image of lungs. Lack of datasets with appropriate number of training images is the major challenge faced by these automated models. In this paper, we propose a model called PneumoGAN that not only augments the training dataset by generating enough number of chest X-ray images from random noise but also has the ability to detect pneumonia from a previously unseen image. The proposed model is inspired from Generative Adversarial Networks (GANs). The discriminator of the proposed PneumoGAN model involves five layers while the generator has six layers in it. The experimental results demonstrate the fact that PneumoGAN has precision, recall and F1 score of 87.71%, 91.4% and 89.52% respectively on benchmark datasets. Moreover, an AUC value of 85% is yielded by the proposed approach. Hence, the proposed model helps doctors to speed up the diagnosis process and narrowing the time required to determine whether a person is a pneumonia victim.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"2 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131878727","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563286
K. Ansal, Chinchu S. Ragamalika, Chippy Susan Rajan
A combination of UWB technology with MIMO technology is applied for designing and simulating a 18×17.5×1.6mm3 sized single element antenna and the two different configurations of 18×35×1.6mm3 sized MIMO array that can be utilized for Ultra Wide Band applications. The single element antenna consists of a compact patch with two rectangular slots which is fed by a 50Ω coplanar waveguide line with a common ground. Having better resonance and a gain of 11dB it achieves a 3 to 10.3GHz wide band. By placing two such single elements on the same side and by making one of them perpendicular to the other, two MIMO antenna arrays are created. Both the arrays almost achieve a wide bandwidth from 3 to 10.3GHz with appreciable resonance and gain of 12.5 and 12.8dB respectively. The designing and simulation is done in ANSYS HFSS version 2021 and the parameters like return loss, radiation pattern, gain, current distribution, VSWR, impedance characteristics, ECC etc. are plotted for their performance comparison.
{"title":"A Two Element CPW Fed MIMO Array for UWB Applications","authors":"K. Ansal, Chinchu S. Ragamalika, Chippy Susan Rajan","doi":"10.1109/ACCESS51619.2021.9563286","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563286","url":null,"abstract":"A combination of UWB technology with MIMO technology is applied for designing and simulating a 18×17.5×1.6mm3 sized single element antenna and the two different configurations of 18×35×1.6mm3 sized MIMO array that can be utilized for Ultra Wide Band applications. The single element antenna consists of a compact patch with two rectangular slots which is fed by a 50Ω coplanar waveguide line with a common ground. Having better resonance and a gain of 11dB it achieves a 3 to 10.3GHz wide band. By placing two such single elements on the same side and by making one of them perpendicular to the other, two MIMO antenna arrays are created. Both the arrays almost achieve a wide bandwidth from 3 to 10.3GHz with appreciable resonance and gain of 12.5 and 12.8dB respectively. The designing and simulation is done in ANSYS HFSS version 2021 and the parameters like return loss, radiation pattern, gain, current distribution, VSWR, impedance characteristics, ECC etc. are plotted for their performance comparison.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132798044","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}
Due to recent circumstances of the pandemic, online platforms are becoming more and more essential for communication in many sectors. But because of this, a lot of negativity and toxic comments are surfacing, resulting in degradation and online abuse. Educational systems and Institutions heavily rely on such platforms for e-learning leading to unrestricted attacks of toxic and negative comments towards teachers and students. Due to this work, issues of constant bullying and online abuse will be reduced. The comments classified are according to the parameters from our self-prepared dataset combined with Kaggle's toxic comment dataset, named as toxic, severely toxic, obscene, threat, insult, and identity hate. Machine Learning algorithms such as Logistic Regression, Random Forest, and Multinomial Naive Bayes are used. For data evaluation, ROC and Hamming scores are used. The output will be shown as the rate of each category in percentile and in a graphical format. This work will help reduce the online bullying and harassment faced by teachers and students and help create a non-toxic learning environment. In this way, the main focus will be on studying and not getting de-motivated and discouraged by hateful comments and people commenting toxic comments will also get reduced.
{"title":"Toxic Comment Analysis for Online Learning","authors":"Manaswi Vichare, Sakshi Thorat, Cdt. Saiba Uberoi, Sheetal Khedekar, S. Jaikar","doi":"10.1109/ACCESS51619.2021.9563344","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563344","url":null,"abstract":"Due to recent circumstances of the pandemic, online platforms are becoming more and more essential for communication in many sectors. But because of this, a lot of negativity and toxic comments are surfacing, resulting in degradation and online abuse. Educational systems and Institutions heavily rely on such platforms for e-learning leading to unrestricted attacks of toxic and negative comments towards teachers and students. Due to this work, issues of constant bullying and online abuse will be reduced. The comments classified are according to the parameters from our self-prepared dataset combined with Kaggle's toxic comment dataset, named as toxic, severely toxic, obscene, threat, insult, and identity hate. Machine Learning algorithms such as Logistic Regression, Random Forest, and Multinomial Naive Bayes are used. For data evaluation, ROC and Hamming scores are used. The output will be shown as the rate of each category in percentile and in a graphical format. This work will help reduce the online bullying and harassment faced by teachers and students and help create a non-toxic learning environment. In this way, the main focus will be on studying and not getting de-motivated and discouraged by hateful comments and people commenting toxic comments will also get reduced.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131089374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563292
Apoorva Shete, H. Soni, Zen Sajnani, Aishwarya Shete
Newspapers and radios are the things of the past, the current generation depends on the internet, specifically social media platforms to stay up to date with the global news. The ease of access, affordability and widespread audience has made these platforms a perfect choice to reach the world. While this has sped up and streamlined news consumption, it is not without drawbacks. The major issue is the proliferation of false/fake news which can have serious repercussions in sensitive matters. Understanding the difference and authenticity of the news is becoming complicated everyday. Social media platforms and online newsletters are responsible for the spread of fake news. However, this problem can be tackled using machine learning techniques and give verifiable news. The paper identifies counterfeit news using Logistic Regression. This model successfully labels a said article as “fake” or “real” with up to 80% accuracy. The paper ends with a review of the model's feasibility and how it would be useful as an impactful mining method as well as proposes the scope of future improvements in the model which will help achieve greater accuracy in the prediction results.
{"title":"Fake News Detection Using Natural Language Processing and Logistic Regression","authors":"Apoorva Shete, H. Soni, Zen Sajnani, Aishwarya Shete","doi":"10.1109/ACCESS51619.2021.9563292","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563292","url":null,"abstract":"Newspapers and radios are the things of the past, the current generation depends on the internet, specifically social media platforms to stay up to date with the global news. The ease of access, affordability and widespread audience has made these platforms a perfect choice to reach the world. While this has sped up and streamlined news consumption, it is not without drawbacks. The major issue is the proliferation of false/fake news which can have serious repercussions in sensitive matters. Understanding the difference and authenticity of the news is becoming complicated everyday. Social media platforms and online newsletters are responsible for the spread of fake news. However, this problem can be tackled using machine learning techniques and give verifiable news. The paper identifies counterfeit news using Logistic Regression. This model successfully labels a said article as “fake” or “real” with up to 80% accuracy. The paper ends with a review of the model's feasibility and how it would be useful as an impactful mining method as well as proposes the scope of future improvements in the model which will help achieve greater accuracy in the prediction results.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"6 2","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"121010529","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563282
Reshma Sajeev, M. G. Krishnan, T. S. Kumar, S. Ashok
Visual servoing is the method of controlling a robot using image input from one or more image sensors to complete a predefined task. This paper examines the effectiveness of a Recurrent Neural Network (RNN) to predict the position and orientation (pose) of an industrial robot manipulator for automatic pick and place applications mainly in unstructured environment. The robot manipulator moves to the target object based on the pose commands obtained from the trained neural network. Various images obtained from the camera attached to the end-effector and corresponding pose of the end-effector are the input and the output data for training the neural network. The performance of the RNN in predicting the robot pose is compared with the feedforward neural (FFN) network and cascade forward neural (CFN) network. The proposed method is validated experimentally using ABB IRB 1200 6-DOF industrial robot manipulator.
{"title":"Design and Implementation of a Robot Pose Predicting Recurrent Neural Network for Visual Servoing Application","authors":"Reshma Sajeev, M. G. Krishnan, T. S. Kumar, S. Ashok","doi":"10.1109/ACCESS51619.2021.9563282","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563282","url":null,"abstract":"Visual servoing is the method of controlling a robot using image input from one or more image sensors to complete a predefined task. This paper examines the effectiveness of a Recurrent Neural Network (RNN) to predict the position and orientation (pose) of an industrial robot manipulator for automatic pick and place applications mainly in unstructured environment. The robot manipulator moves to the target object based on the pose commands obtained from the trained neural network. Various images obtained from the camera attached to the end-effector and corresponding pose of the end-effector are the input and the output data for training the neural network. The performance of the RNN in predicting the robot pose is compared with the feedforward neural (FFN) network and cascade forward neural (CFN) network. The proposed method is validated experimentally using ABB IRB 1200 6-DOF industrial robot manipulator.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"57 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"126280951","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563338
H. Shihabudeen, J. Rajeesh
Collecting salient and relevant information from many images and merging this to generate a quality image is the main goal of image fusion technique. Because of the camera's characteristics while photographing a scene, multi focus images will be produced. Each image of the scene has a different set of features and the merging leads to a good capture of the scene. Activity level measurement and fusion strategy are the critical areas of study in multi focus fusion. To find various focused information in transformed and spatial domains, there have been a lot of algorithms developed. Convolutional neural networks are excellent at representing deep features in an easier format and this property is used to represent multi focus images. Each pixel's activity map is used as a parameter in the fusion strategy. Euclidian norm are a good tool to find the similarities between a set of values. ℓ2 Euclidian norm along with activity map performs the fusion of feature maps collected by residual network. When compared to other fusion algorithms, the presented technique is efficient and improves the image quality. The merged images correlate with human visual perception. The algorithm is suitable for applications like remote sensing, surveillance, and medical diagnosis, etc.
{"title":"Euclidian Norm Based Fusion Strategy for Multi Focus Images","authors":"H. Shihabudeen, J. Rajeesh","doi":"10.1109/ACCESS51619.2021.9563338","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563338","url":null,"abstract":"Collecting salient and relevant information from many images and merging this to generate a quality image is the main goal of image fusion technique. Because of the camera's characteristics while photographing a scene, multi focus images will be produced. Each image of the scene has a different set of features and the merging leads to a good capture of the scene. Activity level measurement and fusion strategy are the critical areas of study in multi focus fusion. To find various focused information in transformed and spatial domains, there have been a lot of algorithms developed. Convolutional neural networks are excellent at representing deep features in an easier format and this property is used to represent multi focus images. Each pixel's activity map is used as a parameter in the fusion strategy. Euclidian norm are a good tool to find the similarities between a set of values. ℓ2 Euclidian norm along with activity map performs the fusion of feature maps collected by residual network. When compared to other fusion algorithms, the presented technique is efficient and improves the image quality. The merged images correlate with human visual perception. The algorithm is suitable for applications like remote sensing, surveillance, and medical diagnosis, etc.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"1 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"131361123","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2021-09-02DOI: 10.1109/ACCESS51619.2021.9563316
N. Archana, K. Hareesh
In computer vision-based applications, the recognition of human activity is always a standard problem. Nowadays, activity recognition is more possible and accurate due to good development in artificial neural networks like convolutional neural network CNN. In many recent works, the recognition model architecture use CNN and long short-term memory units (LSTM) - attention models to extract spatial and temporal features from the input video. This particular work is related to real-time human activity recognition by Resnet and 3D CNN without the involvement of the LSTM- attention model. Here the 2D Resnet is modified to 3D CNN to achieve better human activity recognition accuracy. The wide range of data information from the kinetics dataset can avoid overfitting issues during the training period. And the combination of Resnet and 3D CNN can enhance the accuracy of recognition. As a consequence, a method for detecting, monitoring, and recognizing real-time human motion has been developed.
{"title":"Real-time Human Activity Recognition Using ResNet and 3D Convolutional Neural Networks","authors":"N. Archana, K. Hareesh","doi":"10.1109/ACCESS51619.2021.9563316","DOIUrl":"https://doi.org/10.1109/ACCESS51619.2021.9563316","url":null,"abstract":"In computer vision-based applications, the recognition of human activity is always a standard problem. Nowadays, activity recognition is more possible and accurate due to good development in artificial neural networks like convolutional neural network CNN. In many recent works, the recognition model architecture use CNN and long short-term memory units (LSTM) - attention models to extract spatial and temporal features from the input video. This particular work is related to real-time human activity recognition by Resnet and 3D CNN without the involvement of the LSTM- attention model. Here the 2D Resnet is modified to 3D CNN to achieve better human activity recognition accuracy. The wide range of data information from the kinetics dataset can avoid overfitting issues during the training period. And the combination of Resnet and 3D CNN can enhance the accuracy of recognition. As a consequence, a method for detecting, monitoring, and recognizing real-time human motion has been developed.","PeriodicalId":409648,"journal":{"name":"2021 2nd International Conference on Advances in Computing, Communication, Embedded and Secure Systems (ACCESS)","volume":"75 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2021-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132040678","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：