Pub Date : 2022-05-15DOI: 10.1109/SIU55565.2022.9864744
Enes Colpan, Abdulmajid A.H.A. Mohammed, Ö. N. Gerek
The great success of deep learning methods for object detection rendered such methods the fundamental choice in related applications. Popular choices for multiple object detection in video sequences include convolutional neural networks, such as YOLO, MobileNet-SSD and Faster R-CNN, which typically split image frames to small rectangular regions and attempts to find bounding boxes of sought–after objects. Current research of such methods mostly focus on speeding–up the implementations or improving the network layers’ learning properties. As a new approach, this work appends a simple post processing stage at the end of such networks to reinforce decision robustness using a sequential decision process through sequential video frames. The sequential frames provide a better confidence on the existence of an object, when a probable object was also estimated in the previous frame. Once the confidence level overshoots a predetermined threshold, objects that are difficult to be detected in a single frame get accurately detected.
{"title":"Object Recognition with Sequential Decision Reinforcement of Deep Learning","authors":"Enes Colpan, Abdulmajid A.H.A. Mohammed, Ö. N. Gerek","doi":"10.1109/SIU55565.2022.9864744","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864744","url":null,"abstract":"The great success of deep learning methods for object detection rendered such methods the fundamental choice in related applications. Popular choices for multiple object detection in video sequences include convolutional neural networks, such as YOLO, MobileNet-SSD and Faster R-CNN, which typically split image frames to small rectangular regions and attempts to find bounding boxes of sought–after objects. Current research of such methods mostly focus on speeding–up the implementations or improving the network layers’ learning properties. As a new approach, this work appends a simple post processing stage at the end of such networks to reinforce decision robustness using a sequential decision process through sequential video frames. The sequential frames provide a better confidence on the existence of an object, when a probable object was also estimated in the previous frame. Once the confidence level overshoots a predetermined threshold, objects that are difficult to be detected in a single frame get accurately detected.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"2021 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117176639","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864747
Ferda Gul Aydin Emekligil, Ilkay Öksüz
Designing visual content and characters for games is a time consuming task even for designers and illustrators with experience. Most of the game companies and developers use procedural methods to automate the design process. The visual content produced by these algorithms is limited in terms of variation. In this paper, we propose to use Generative Adversarial Networks (GANs) for visual content production. Two different rpg and dnd visual image datasets were collected over the internet for training and 6 different GAN models were trained on them. In 3 of 18 experiments, transfer learning methods are used because of the limited datasets. The Frechet Inception Distance metric was used to compare the model results. As a result, SNGAN was the most successful in both datasets. Moreover, the transfer learning method (WGAN-GP, BigGAN) was more successful than the from scratch method.
{"title":"Game Character Generation with Generative Adversarial Networks","authors":"Ferda Gul Aydin Emekligil, Ilkay Öksüz","doi":"10.1109/SIU55565.2022.9864747","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864747","url":null,"abstract":"Designing visual content and characters for games is a time consuming task even for designers and illustrators with experience. Most of the game companies and developers use procedural methods to automate the design process. The visual content produced by these algorithms is limited in terms of variation. In this paper, we propose to use Generative Adversarial Networks (GANs) for visual content production. Two different rpg and dnd visual image datasets were collected over the internet for training and 6 different GAN models were trained on them. In 3 of 18 experiments, transfer learning methods are used because of the limited datasets. The Frechet Inception Distance metric was used to compare the model results. As a result, SNGAN was the most successful in both datasets. Moreover, the transfer learning method (WGAN-GP, BigGAN) was more successful than the from scratch method.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"30 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"128236864","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864707
Yasar Tekin, A. Cosar
Topic modeling is a text mining technique used for automatic extraction of topics addressed in document collections. Although there are different topic models proposed by researchers, the most preferred one is Latent Dirichlet Allocation (LDA). Despite such widespread use, uncertainties about LDA have not been fully resolved yet. In this study, the effect of using abstracts or full-text articles on LDA model parameters is investigated. For this purpose, LDA parameters are optimized on abstracts and full-texts of articles published in two different scientific journals and the results obtained are compared with each other.
{"title":"Abstract or Full-text in Topic Modeling?","authors":"Yasar Tekin, A. Cosar","doi":"10.1109/SIU55565.2022.9864707","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864707","url":null,"abstract":"Topic modeling is a text mining technique used for automatic extraction of topics addressed in document collections. Although there are different topic models proposed by researchers, the most preferred one is Latent Dirichlet Allocation (LDA). Despite such widespread use, uncertainties about LDA have not been fully resolved yet. In this study, the effect of using abstracts or full-text articles on LDA model parameters is investigated. For this purpose, LDA parameters are optimized on abstracts and full-texts of articles published in two different scientific journals and the results obtained are compared with each other.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"107 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124744287","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864946
Ahmet Saim Yilmaz, H. Çukurtepe, Emin Kugu
Global Positioning Systems and Wi-Fi Positioning Systems are widely used for positioning purposes. Geo-location services are extremely vulnerable to location spoofing attacks and have been researched for quite some time. The majority of the researches have been focused on devices like smartphones, unmanned aerial vehicles, automobiles, and etc. E-scooters have also joined the family of devices having geo-location capabilities. In this study, we give the computation basics of Global Positioning Systems and Wi-Fi Positioning Systems, and analyze the threats for location spoofing attacks on E-scooter Sharing System. It is shown that limiting or unlimiting the speed of e-scooter, preventing users and maintenance crew to find the e-scooter, stopping the device or prevent user’s from ending the ride is possible with location spoofing attacks.
{"title":"Analysis of Location Spoofing Threats on E-Scooter Sharing","authors":"Ahmet Saim Yilmaz, H. Çukurtepe, Emin Kugu","doi":"10.1109/SIU55565.2022.9864946","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864946","url":null,"abstract":"Global Positioning Systems and Wi-Fi Positioning Systems are widely used for positioning purposes. Geo-location services are extremely vulnerable to location spoofing attacks and have been researched for quite some time. The majority of the researches have been focused on devices like smartphones, unmanned aerial vehicles, automobiles, and etc. E-scooters have also joined the family of devices having geo-location capabilities. In this study, we give the computation basics of Global Positioning Systems and Wi-Fi Positioning Systems, and analyze the threats for location spoofing attacks on E-scooter Sharing System. It is shown that limiting or unlimiting the speed of e-scooter, preventing users and maintenance crew to find the e-scooter, stopping the device or prevent user’s from ending the ride is possible with location spoofing attacks.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"7 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129491683","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}
In Italy, change in the voters’ behaviours of obtaining information about politics is observed from 2001 to 2019. Particularly, a continuous decrease in the number of voters between 2006 and 2018, is achieved. This study investigates whether acquiring political information from social networks (e.g. friends and relatives) and mass media (e.g. tv and radio) is related to the decision of people to vote or not. Linear regression analysis is applied to discover the relationship between turnout rate and political means. It is found that getting informed about politics from relatives, acquaintances, political organizations, trade unions, radio and weekly magazines are not statistically meaningful to explain changes in turnout rates. Friends have an impact on turnout rates in a negative direction. Besides, getting information from tv and newspapers has a positive impact on turnout rates. It is also observed that mass media is more effective than social networks on turnout rates.
{"title":"Effect of Social Network and Mass Media on Turnout Rates in Italy","authors":"Cansu Damla Yilmaz, Safa Nur Altuncu Kaan, Sumeyye Agac, Didem Gündoğdu","doi":"10.1109/SIU55565.2022.9864757","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864757","url":null,"abstract":"In Italy, change in the voters’ behaviours of obtaining information about politics is observed from 2001 to 2019. Particularly, a continuous decrease in the number of voters between 2006 and 2018, is achieved. This study investigates whether acquiring political information from social networks (e.g. friends and relatives) and mass media (e.g. tv and radio) is related to the decision of people to vote or not. Linear regression analysis is applied to discover the relationship between turnout rate and political means. It is found that getting informed about politics from relatives, acquaintances, political organizations, trade unions, radio and weekly magazines are not statistically meaningful to explain changes in turnout rates. Friends have an impact on turnout rates in a negative direction. Besides, getting information from tv and newspapers has a positive impact on turnout rates. It is also observed that mass media is more effective than social networks on turnout rates.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"16 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"129929727","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864781
Tunahan Bozkan, Tuna Çakar, A. Sayar, Seyit Ertugrul
In this study, it is aimed to predict whether customers operating in the factoring sector will continue to trade in the next three months after the last transaction date, using data- driven machine learning models, based on their past transaction movements and their risk, limit and company data. As a result of the models established, Loss Analysis (Churn) of two different customer groups (Real and Legal factory) wascarried out. It was estimated by the XGBoost model with anF1 Score of 74% and 77%. Thanks to this modeling, it was aimed to increase the retention rate of customers through special promotions and campaigns to be made to these customer groups, together with the prediction of the customerswho will leave. Thanks to the increase in retention rates, a direct contribution to the transaction volume on a company basis was ensured.
{"title":"Customer Segmentation and Churn Prediction via Customer Metrics","authors":"Tunahan Bozkan, Tuna Çakar, A. Sayar, Seyit Ertugrul","doi":"10.1109/SIU55565.2022.9864781","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864781","url":null,"abstract":"In this study, it is aimed to predict whether customers operating in the factoring sector will continue to trade in the next three months after the last transaction date, using data- driven machine learning models, based on their past transaction movements and their risk, limit and company data. As a result of the models established, Loss Analysis (Churn) of two different customer groups (Real and Legal factory) wascarried out. It was estimated by the XGBoost model with anF1 Score of 74% and 77%. Thanks to this modeling, it was aimed to increase the retention rate of customers through special promotions and campaigns to be made to these customer groups, together with the prediction of the customerswho will leave. Thanks to the increase in retention rates, a direct contribution to the transaction volume on a company basis was ensured.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"123 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"130333004","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864749
Dogukan Gozler, Beyazit Isik, C. Topal
In daily life, people spend a significant part of their time waiting in queues at numerous places such as banks, airports, cafeterias and market cash registers. Various queue analysis and management tools have been developed to reduce this wasted time. One of these tools is the systems that analyze the queues and calculate the average waiting time. In this study, a computer vision method that calculates the average waiting time by detecting and tracking the people waiting in the queue is proposed. Thanks to this method, it is aimed that people can see how long they will wait before queuing. In the developed method, people waiting in a direction were analyzed by using object detection methods, and the times of joining and leaving the queue were tried to be determined. Due to the high processing load of the object detection algorithms, object tracking algorithms are used so that the method can work in real-time. The method developed according to the experimental studies can process the 640×480 resolution video on a mid-level GPU with %88.46 accuracy and speeds up to 95.51 fps.
{"title":"A Real-time Queue Tracking Method for Waiting Time Estimation","authors":"Dogukan Gozler, Beyazit Isik, C. Topal","doi":"10.1109/SIU55565.2022.9864749","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864749","url":null,"abstract":"In daily life, people spend a significant part of their time waiting in queues at numerous places such as banks, airports, cafeterias and market cash registers. Various queue analysis and management tools have been developed to reduce this wasted time. One of these tools is the systems that analyze the queues and calculate the average waiting time. In this study, a computer vision method that calculates the average waiting time by detecting and tracking the people waiting in the queue is proposed. Thanks to this method, it is aimed that people can see how long they will wait before queuing. In the developed method, people waiting in a direction were analyzed by using object detection methods, and the times of joining and leaving the queue were tried to be determined. Due to the high processing load of the object detection algorithms, object tracking algorithms are used so that the method can work in real-time. The method developed according to the experimental studies can process the 640×480 resolution video on a mid-level GPU with %88.46 accuracy and speeds up to 95.51 fps.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"72 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123846652","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864930
Mustafa Kagan Gürkan, N. Arica, F. Yarman-Vural
In this paper, we propose a scoring algorithm for measuring the interpretability of CNN models by focusing on the feature extraction operation at the convolutional layers. The proposed approach is based on the principal of concept analysis, for a predefined list of concepts. A map of the network is created based on its responsiveness against each concept. Once this map is ready, various images can be applied as inputs and they are matched with the concepts whose hidden nodes are highly activated. Finally, the evaluation algorithm kicks in to use these descriptions during the final prediction and provides human-understandable explanations.
{"title":"A Scoring Method for Interpretability of Concepts in Convolutional Neural Networks","authors":"Mustafa Kagan Gürkan, N. Arica, F. Yarman-Vural","doi":"10.1109/SIU55565.2022.9864930","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864930","url":null,"abstract":"In this paper, we propose a scoring algorithm for measuring the interpretability of CNN models by focusing on the feature extraction operation at the convolutional layers. The proposed approach is based on the principal of concept analysis, for a predefined list of concepts. A map of the network is created based on its responsiveness against each concept. Once this map is ready, various images can be applied as inputs and they are matched with the concepts whose hidden nodes are highly activated. Finally, the evaluation algorithm kicks in to use these descriptions during the final prediction and provides human-understandable explanations.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"46 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123522433","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 : 2022-05-15DOI: 10.1109/SIU55565.2022.9864938
Emre Bulbul, G. Akar
Visual field testing is the gold standard for evaluating a patient’s visual field. Visual field testing is required for monitoring and diagnosis of several disorders, including glaucoma, which affects more than 80 million individuals. While the patient is fixated at a certain place, light of various luminosities is sent to fixed locations, and the sensitivities to light at each position are calculated by recording the patient’s responses to observed stimuli. Virtual reality headsets have just begun to be used to conduct visual field assessments due to their design and digital displays. However, because the testing takes so long, patients become fatigued, which reduces cooperation and test accuracy. It also restricts the number of tests a clinic may do in a single day. The number of testable point locations is expanded using a digital screen in this article, and the effect of selecting an optimal subset of sites, which is discovered using a reinforcement learning approach to reduce test length, is studied. In addition, the impact of employing predicted future visual field test results in testing on the test time is compared to traditional testing procedures.
{"title":"The Effect of Virtual Reality and Prediction in Visual Field Test","authors":"Emre Bulbul, G. Akar","doi":"10.1109/SIU55565.2022.9864938","DOIUrl":"https://doi.org/10.1109/SIU55565.2022.9864938","url":null,"abstract":"Visual field testing is the gold standard for evaluating a patient’s visual field. Visual field testing is required for monitoring and diagnosis of several disorders, including glaucoma, which affects more than 80 million individuals. While the patient is fixated at a certain place, light of various luminosities is sent to fixed locations, and the sensitivities to light at each position are calculated by recording the patient’s responses to observed stimuli. Virtual reality headsets have just begun to be used to conduct visual field assessments due to their design and digital displays. However, because the testing takes so long, patients become fatigued, which reduces cooperation and test accuracy. It also restricts the number of tests a clinic may do in a single day. The number of testable point locations is expanded using a digital screen in this article, and the effect of selecting an optimal subset of sites, which is discovered using a reinforcement learning approach to reduce test length, is studied. In addition, the impact of employing predicted future visual field test results in testing on the test time is compared to traditional testing procedures.","PeriodicalId":115446,"journal":{"name":"2022 30th Signal Processing and Communications Applications Conference (SIU)","volume":"166 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2022-05-15","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"116308997","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：