Pub Date : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633793
V. Abhijith, Chandana Phanidhar Sai Sravan, D. Raju, T. Sasikala
With spam filtering techniques have been improved in social websites like G mail., spammers find their place in other famous social platforms like Twitter, Facebook. Therefore, an effective spam filtering technology is essential for platforms like Twitter, Facebook, etc. We have developed a web application that will be able to find out whether a particular tweet from Twitter is malicious or non- malicious based on the Url that the tweet possesses by considering both text-based and Url-based features. We have employed machine learning techniques to classify the tweet content after preprocessing the data that we have fetched from Twitter with the help of tokens that we obtain after creating the Twitter developer account. We are classifying a tweet based on five different features, these features can be most commonly found in malicious tweets as per our research. The results that are obtained from our experiment show that our approach could efficiently identify malicioustweets.
{"title":"Detection of Malicious URLs in Twitter","authors":"V. Abhijith, Chandana Phanidhar Sai Sravan, D. Raju, T. Sasikala","doi":"10.1109/ICSES52305.2021.9633793","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633793","url":null,"abstract":"With spam filtering techniques have been improved in social websites like G mail., spammers find their place in other famous social platforms like Twitter, Facebook. Therefore, an effective spam filtering technology is essential for platforms like Twitter, Facebook, etc. We have developed a web application that will be able to find out whether a particular tweet from Twitter is malicious or non- malicious based on the Url that the tweet possesses by considering both text-based and Url-based features. We have employed machine learning techniques to classify the tweet content after preprocessing the data that we have fetched from Twitter with the help of tokens that we obtain after creating the Twitter developer account. We are classifying a tweet based on five different features, these features can be most commonly found in malicious tweets as per our research. The results that are obtained from our experiment show that our approach could efficiently identify malicioustweets.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"30 1","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"91090801","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633951
Arpan Ghoshal, Rohan Kamath, D. Uma
In this modern age of information, the answer to any question is readily available on the internet. The chronology of the solution to open-domain question-answering capabilities is primitive manual question mining, followed by search engine revolution. Later, intelligent assistants became a robust method, which is being used for a long time. However, it has been noticed that the intelligent assistant responds to simple questions but not to the complicated ones that demand comprehension of the question's abstract. This paper introduces an innovative, dynamic technique called CATAQ, i.e., Concise Answer to Any Question that responds to complicated questions by providing accurate answers. Furthermore, CATAQ introduces a solution for dynamic information retrieval, which involves the selection of the most appropriate webpage links, based on the link parsing the content of the webpage irrespective of its structure, pre-processing the content, extracting the pertinent information with semantic search and summarizing by prioritizing question's keywords. It is observed that CAT AQ outperformed the present intelligent assistants while answering complex questions which include anunusual and large variety of topics and keeping the performance consistent while answering simple questions. Concluding, CATAQ can answer any question concisely, assuming the related information is available on the internet.
{"title":"CATAQ: Concise Answer to any Question","authors":"Arpan Ghoshal, Rohan Kamath, D. Uma","doi":"10.1109/ICSES52305.2021.9633951","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633951","url":null,"abstract":"In this modern age of information, the answer to any question is readily available on the internet. The chronology of the solution to open-domain question-answering capabilities is primitive manual question mining, followed by search engine revolution. Later, intelligent assistants became a robust method, which is being used for a long time. However, it has been noticed that the intelligent assistant responds to simple questions but not to the complicated ones that demand comprehension of the question's abstract. This paper introduces an innovative, dynamic technique called CATAQ, i.e., Concise Answer to Any Question that responds to complicated questions by providing accurate answers. Furthermore, CATAQ introduces a solution for dynamic information retrieval, which involves the selection of the most appropriate webpage links, based on the link parsing the content of the webpage irrespective of its structure, pre-processing the content, extracting the pertinent information with semantic search and summarizing by prioritizing question's keywords. It is observed that CAT AQ outperformed the present intelligent assistants while answering complex questions which include anunusual and large variety of topics and keeping the performance consistent while answering simple questions. Concluding, CATAQ can answer any question concisely, assuming the related information is available on the internet.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"101 1","pages":"1-8"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90266034","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633918
Heth Gala, Jenish Hirpara, Mihir Shah, Jash Shah, L. D'mello
The recent survey presents the increase in people with cognitive impairment and disorders related to the same. The current project aims to reduce the distress and impairment in social and academic functioning caused by the symptoms of various disorders such as dyslexia, dysgraphia, dysphasia, visual processing orders, and others, by adopting a model that tackles difficulties associated with various cognitive capacities as well as auditory and visual impairments. The first section of the three-phase mobile application converts the input image of scripts to concise, visual representations alongside a speech assistant that restates the summarized version of the text. The second phase has been designed to assist people with difficulties in grasping long audio notes, by converting them into a set of intuitive visual representations with speed and volume regulated speech assistant. Furthermore, the model has been extended to creating a tutorial environment on the app to improve the syntactic spelling of general words in the English language for people with dyslexia. Intending to cater to the needs of the user with all three sections in a single cross-platform application, we believe that we were successful in creating the assistant as envisioned. The appropriate usability of the app for the target audience presented above shall mark the success of the app as well as the AI community working for years in this field.
{"title":"Visual and Auditory Assistant for people with various cognitive impairments","authors":"Heth Gala, Jenish Hirpara, Mihir Shah, Jash Shah, L. D'mello","doi":"10.1109/ICSES52305.2021.9633918","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633918","url":null,"abstract":"The recent survey presents the increase in people with cognitive impairment and disorders related to the same. The current project aims to reduce the distress and impairment in social and academic functioning caused by the symptoms of various disorders such as dyslexia, dysgraphia, dysphasia, visual processing orders, and others, by adopting a model that tackles difficulties associated with various cognitive capacities as well as auditory and visual impairments. The first section of the three-phase mobile application converts the input image of scripts to concise, visual representations alongside a speech assistant that restates the summarized version of the text. The second phase has been designed to assist people with difficulties in grasping long audio notes, by converting them into a set of intuitive visual representations with speed and volume regulated speech assistant. Furthermore, the model has been extended to creating a tutorial environment on the app to improve the syntactic spelling of general words in the English language for people with dyslexia. Intending to cater to the needs of the user with all three sections in a single cross-platform application, we believe that we were successful in creating the assistant as envisioned. The appropriate usability of the app for the target audience presented above shall mark the success of the app as well as the AI community working for years in this field.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"1 1","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"76558703","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633891
D. N. Yethindra, G. Deepak
Due to the increased prevalence of web recommendation systems after years of research, it has unarguably become the ultimate solution for efficient functioning of any e-commerce or user supportive digital domain. Though a variety of algorithms have been tested to meet the expectations of users in order to be decision supportive, this paper proposes a potential framework for recommendation of men's clothing. The focus of the system is to improve the efficiency of the recommendation to cope up to the speed of the user's thought process and expectations at the same time generate only those options that have been validated closely to the user's style hunt trajectory. In the presented approach the user's historical click data and searches is preprocessed and converted into query words. The features are extracted from the on ontology of fashion with the help of query words. The ontology used in this paper is highly domain specific. External sources such as fashion reviews, fashion e-magazines, fashion blogs and fashion trends from e-commerce websites are converted into query words and used for feature enrichment. The dataset is provided for classification using logistic regression, and only the top 50% of results from the classification undergoes semantic similarity computation. Normalized google distance and SemantoSim measure are the methods used for emantic similarity computation, this happens mainly for the relevance of the results. The recommendations of fashion items and fashion brands are suggested to the user based on the results gotten from semantic similarity. The accuracy of the Onto infused recommendation system is 97.14% and average precision is 96.31%.
{"title":"A Semantic Approach for Fashion Recommendation Using Logistic Regression and Ontologies","authors":"D. N. Yethindra, G. Deepak","doi":"10.1109/ICSES52305.2021.9633891","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633891","url":null,"abstract":"Due to the increased prevalence of web recommendation systems after years of research, it has unarguably become the ultimate solution for efficient functioning of any e-commerce or user supportive digital domain. Though a variety of algorithms have been tested to meet the expectations of users in order to be decision supportive, this paper proposes a potential framework for recommendation of men's clothing. The focus of the system is to improve the efficiency of the recommendation to cope up to the speed of the user's thought process and expectations at the same time generate only those options that have been validated closely to the user's style hunt trajectory. In the presented approach the user's historical click data and searches is preprocessed and converted into query words. The features are extracted from the on ontology of fashion with the help of query words. The ontology used in this paper is highly domain specific. External sources such as fashion reviews, fashion e-magazines, fashion blogs and fashion trends from e-commerce websites are converted into query words and used for feature enrichment. The dataset is provided for classification using logistic regression, and only the top 50% of results from the classification undergoes semantic similarity computation. Normalized google distance and SemantoSim measure are the methods used for emantic similarity computation, this happens mainly for the relevance of the results. The recommendations of fashion items and fashion brands are suggested to the user based on the results gotten from semantic similarity. The accuracy of the Onto infused recommendation system is 97.14% and average precision is 96.31%.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"48 1","pages":"1-6"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"74760254","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633858
Maturi Tanuj, Aishwarya Virigineni, Apoorva Mani, R. Subramani
The paper describes a comparative study of three different approaches that are used for image blending. The main focus will remain on the approaches where single source image and target is composed in the gradient domain. The main aim of the study is to portray the importance of considering gradients in image blending and why it makes the blending more realistic and effective. The study of all the approaches that are taken under consideration have been executed using MATLAB. The three different approaches are Naive blending, poisson image blending and Mixed gradient approach of image blending.
{"title":"Comparative Study of Gradient Domain Based Image Blending Approaches","authors":"Maturi Tanuj, Aishwarya Virigineni, Apoorva Mani, R. Subramani","doi":"10.1109/ICSES52305.2021.9633858","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633858","url":null,"abstract":"The paper describes a comparative study of three different approaches that are used for image blending. The main focus will remain on the approaches where single source image and target is composed in the gradient domain. The main aim of the study is to portray the importance of considering gradients in image blending and why it makes the blending more realistic and effective. The study of all the approaches that are taken under consideration have been executed using MATLAB. The three different approaches are Naive blending, poisson image blending and Mixed gradient approach of image blending.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"132 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"79635933","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633961
C. Dadiyala, Asha Ambhaikar
In Stock Prediction, the aim is to predict future stock values with desirable accuracy. Our research aims to offer a method for technical analysis of pattern based stock prediction using Machine Learning on the historical stock data. The newly designed method is based on GA with the appropriate modifications needed for the prediction. We have performed various experiments using the historical data of a few companies and the results confirmed the accuracy and efficiency of the system as it is generating promising predictions. This designed model executes a prediction process that is not influenced by any other external factors.
{"title":"Technical Analysis of Pattern Based Stock Prediction Model Using Machine Learning","authors":"C. Dadiyala, Asha Ambhaikar","doi":"10.1109/ICSES52305.2021.9633961","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633961","url":null,"abstract":"In Stock Prediction, the aim is to predict future stock values with desirable accuracy. Our research aims to offer a method for technical analysis of pattern based stock prediction using Machine Learning on the historical stock data. The newly designed method is based on GA with the appropriate modifications needed for the prediction. We have performed various experiments using the historical data of a few companies and the results confirmed the accuracy and efficiency of the system as it is generating promising predictions. This designed model executes a prediction process that is not influenced by any other external factors.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"69 1","pages":"1-9"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"80281749","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633881
Sajeel Mehta, D. Pawade, Yash Nayyar, Irfan A. Siddavatam, Anoop Tiwart, A. Dalvi
Honeypots are recent innovation in intrusion detection technology. They are the traps designed to basically entrap potential intruders and log their activities. The main objective of such systems is to collect the information about the intruders, deviate them from accessing critical systems, push them to stay on top of the system for some time so their behavior can be observed. We have used Cowrie Honeypot to achieve the above objectives. The log of intruder activities is maintained which is processed and graphically visualized using ELK. This intruder activity data is useful to know the intruder behavior and accordingly safety measures can be employed against that. In extension to data visualization, we have also implemented the probabilistic approach to predict the directory traverser pattern of the intruder. This will help us to understand the next traverser step in advance so that we can take precautionary measures to avoid it.
{"title":"Cowrie Honeypot Data Analysis and Predicting the Directory Traverser Pattern during the Attack","authors":"Sajeel Mehta, D. Pawade, Yash Nayyar, Irfan A. Siddavatam, Anoop Tiwart, A. Dalvi","doi":"10.1109/ICSES52305.2021.9633881","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633881","url":null,"abstract":"Honeypots are recent innovation in intrusion detection technology. They are the traps designed to basically entrap potential intruders and log their activities. The main objective of such systems is to collect the information about the intruders, deviate them from accessing critical systems, push them to stay on top of the system for some time so their behavior can be observed. We have used Cowrie Honeypot to achieve the above objectives. The log of intruder activities is maintained which is processed and graphically visualized using ELK. This intruder activity data is useful to know the intruder behavior and accordingly safety measures can be employed against that. In extension to data visualization, we have also implemented the probabilistic approach to predict the directory traverser pattern of the intruder. This will help us to understand the next traverser step in advance so that we can take precautionary measures to avoid it.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"58 1","pages":"1-4"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"86893391","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633797
G. Shobana, S. Bushra
The number of people affected due to Cardiovascular diseases has escalated in recent years. The sedentary lifestyle, certain genetic factors, obesity, lack of exercise and stressful work environments act as a catalyst in the progress of the disease. Heart failure is one of the Cardio-vascular diseases that occur due to improper flow of blood and inadequate level of oxygen in the blood. Researchers apply machine learning algorithms to identify the crucial factors involved in heart diseases. The data obtained from patients are explored and analyzed using various data mining tools to derive relevant and accurate outcomes. In this paper, two popular machine learning platforms Scikit-Learn and Orange are investigated by implementing Seven machine learning techniques and Boosting algorithms, their performance on the Heart Failure dataset is explored with various training and testing ratios. Their best training and the testing split are determined. Performance of the datamining tools are examined and various metrics are evaluated. Machine learning techniques like traditional Logistic Regression, Naïve Bayes and ensemble Random Forest models had higher prediction accuracies. The Boosting algorithms performed efficiently than other common models with 89%.
{"title":"Prediction of Cardiovascular Disease using Multiple Machine Learning Platforms","authors":"G. Shobana, S. Bushra","doi":"10.1109/ICSES52305.2021.9633797","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633797","url":null,"abstract":"The number of people affected due to Cardiovascular diseases has escalated in recent years. The sedentary lifestyle, certain genetic factors, obesity, lack of exercise and stressful work environments act as a catalyst in the progress of the disease. Heart failure is one of the Cardio-vascular diseases that occur due to improper flow of blood and inadequate level of oxygen in the blood. Researchers apply machine learning algorithms to identify the crucial factors involved in heart diseases. The data obtained from patients are explored and analyzed using various data mining tools to derive relevant and accurate outcomes. In this paper, two popular machine learning platforms Scikit-Learn and Orange are investigated by implementing Seven machine learning techniques and Boosting algorithms, their performance on the Heart Failure dataset is explored with various training and testing ratios. Their best training and the testing split are determined. Performance of the datamining tools are examined and various metrics are evaluated. Machine learning techniques like traditional Logistic Regression, Naïve Bayes and ensemble Random Forest models had higher prediction accuracies. The Boosting algorithms performed efficiently than other common models with 89%.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"27 1","pages":"1-7"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"87475365","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633852
K. Vinisha, E. Kalpana
The present paper discusses about a smart vehiclewhich has the ability to identify any kind of fault occurrence in the power transmission system. It also gives the risk alerts' to the vehicle driver through LCD and at the same time by using a mobile application. This smart vehicle consists of different sensors which are located at the power transmission system of the vehicle. The sensors values from the vehicle are collected and sent to the controller which is compared with some specified independent values. The compression is done using machine learning algorithms which are very useful for achieving a system with high accuracy. With this design we can even achieve the concept of internet of vehicle (IoV), as we are using GPS to track the vehicle and a mobile application to indicate the risk. Thetest was run and outcome of the test was very effective. With the help of this system we can reduce losses of human life and increase the vehicle life span. To achieve the required system we are using machine learning and python, as they are the recentera high level technologies and provide great accuracy.
{"title":"Effectuation of Machine Learning for Fault Classification on Vehicle Power Transmission System","authors":"K. Vinisha, E. Kalpana","doi":"10.1109/ICSES52305.2021.9633852","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633852","url":null,"abstract":"The present paper discusses about a smart vehiclewhich has the ability to identify any kind of fault occurrence in the power transmission system. It also gives the risk alerts' to the vehicle driver through LCD and at the same time by using a mobile application. This smart vehicle consists of different sensors which are located at the power transmission system of the vehicle. The sensors values from the vehicle are collected and sent to the controller which is compared with some specified independent values. The compression is done using machine learning algorithms which are very useful for achieving a system with high accuracy. With this design we can even achieve the concept of internet of vehicle (IoV), as we are using GPS to track the vehicle and a mobile application to indicate the risk. Thetest was run and outcome of the test was very effective. With the help of this system we can reduce losses of human life and increase the vehicle life span. To achieve the required system we are using machine learning and python, as they are the recentera high level technologies and provide great accuracy.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"130 1","pages":"1-5"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"90920177","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 : 2021-09-24DOI: 10.1109/ICSES52305.2021.9633781
Mohammed Ehsan Ur Rahman, Hrudheeshta Anishetty, Arjun Kumar Kollpaka, Aishwarya Yelishetty, S. Ganta
Our proposed work is a research project that does quantitative analysis of various basic image manipulation techniques as processes for augmentation of image type data on the accuracy of deep learning task of hand-written digit recognition on MNIST dataset. The paper also presents a detailed comparison of various parameters such as computation burden, storage requirements for model storage, accuracy, and loss function value of the results obtained by using basic image manipulation techniques as image data augmentation techniques with those data augmentation mechanisms that are rooted in deep learning. The results that we have obtained on MNIST dataset without data augmentation applied are accuracy of 97.80% and loss of 0.320, whereas the highest accuracy was achieved by adjusting brightness as the data augmentation technique with 98.57% accuracy and 0.301 loss value. In the view of our results, we recommend that basic image manipulation-based data augmentation techniques must be used to address overfitting instead of memory or computationally expensive deep learning-based image augmentation techniques. This strategy also helps enhance the performance of various image data-based deep learning pipelines and makes these models more robust.
{"title":"A Quantitative Analysis of Basic vs. Deep Learning-based Image Data Augmentation Techniques","authors":"Mohammed Ehsan Ur Rahman, Hrudheeshta Anishetty, Arjun Kumar Kollpaka, Aishwarya Yelishetty, S. Ganta","doi":"10.1109/ICSES52305.2021.9633781","DOIUrl":"https://doi.org/10.1109/ICSES52305.2021.9633781","url":null,"abstract":"Our proposed work is a research project that does quantitative analysis of various basic image manipulation techniques as processes for augmentation of image type data on the accuracy of deep learning task of hand-written digit recognition on MNIST dataset. The paper also presents a detailed comparison of various parameters such as computation burden, storage requirements for model storage, accuracy, and loss function value of the results obtained by using basic image manipulation techniques as image data augmentation techniques with those data augmentation mechanisms that are rooted in deep learning. The results that we have obtained on MNIST dataset without data augmentation applied are accuracy of 97.80% and loss of 0.320, whereas the highest accuracy was achieved by adjusting brightness as the data augmentation technique with 98.57% accuracy and 0.301 loss value. In the view of our results, we recommend that basic image manipulation-based data augmentation techniques must be used to address overfitting instead of memory or computationally expensive deep learning-based image augmentation techniques. This strategy also helps enhance the performance of various image data-based deep learning pipelines and makes these models more robust.","PeriodicalId":6777,"journal":{"name":"2021 International Conference on Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES)","volume":"40 1","pages":"1-9"},"PeriodicalIF":0.0,"publicationDate":"2021-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"85100983","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：