Browsing by Author "Karabey Aksakalli, I."

Now showing 1 - 6 of 6

Citation - Scopus: 1
Deep Learning Based Lesion Detection on Dental Panoramic Radiographs
(Institute of Electrical and Electronics Engineers Inc., 2023) Demir, K.; Karabey Aksakalli, I.; Baygin, N.; Sokmen, O.C.
With the development of technology, one of the new breakthroughs in the health field has been dental lesion detection. Dental lesion occurs with tissue changes that occur in the parts of the teeth that are especially decayed or filled, close to the gingiva. In cases where this change is not visible and cannot be distinguished from healthy tissue, computer-aided artificial intelligence methods can be used to assist the dentist quickly and accurately. In this study, panoramic images among radiographic images were obtained and deep learning methods applied to these images were compared in terms of accuracy, sensitivity, precision, F1 score, frame per second and intersection of units metrics. The labels of 660 panoramic images were validated by a specialist dentist and the images were trained using YOLOv5, YOLOv8 and SAM models. The experimental results show that YOLOv8 method reached the highest performance value with 0.975 accuracy, 0.978 precision and 0.986 sensitivity rate in the generated test model. © 2023 IEEE.
Citation - Scopus: 2
Deep Learning Based Mobile Platform for Predicting Passenger Congestion in Public Transportation
(Institute of Electrical and Electronics Engineers Inc., 2024) Kavanoz, A.; Kosak, H.I.; Karabey Aksakalli, I.; Bayindir, L.
Population growth in Turkey has led to significant congestion in transportation systems. This congestion results in delays for individuals using public transportation, often causing them to arrive hours early to ensure punctuality at their destinations. Passengers frequently lack real-time information about the current occupancy of public transportation vehicles, leading to uncertainty and potential inefficiencies as they wait at bus stops. In some cases, public transportation vehicles reach full capacity, leaving passengers at the stop unserved. Moreover, it is crucial for passengers to be aware of the crowd levels at various stops, as this affects their ability to efficiently board and travel, particularly on smaller vehicles such as minibuses where stop density is a key consideration. As a result, passengers must make informed decisions based on both vehicle and stop density, as well as the estimated vehicle arrival time. To address these challenges, we propose a deep learning-based mobile platform designed to detect crowd density at bus stops and within public transportation vehicles. This platform will notify both the transit manager and the passengers about real-time density information. Our approach involves analyzing 20 -minute video footage from Mersin Municipality, Turkey, using Mask-RCNN,YOLOv5, YOLOv8, and YOLOv9 models to count passengers boarding and alighting. The models were evaluated on accuracy, precision, sensitivity, F1 score, Frame Per Second (FPS), and Intersection Over Union (IOU) metrics. The experimental results indicated that the YOLOv9 model significantly outperformed the other models, achieving 88 % accuracy, 86.2 % precision, 89 % recall, 15.2 fps, and 0.78 IoU. © 2024 IEEE.
An Evaluation of Fingerprint-Based Indoor Localization Techniques
(Institute of Electrical and Electronics Engineers Inc., 2015) Karabey Aksakalli, I.; Bayindir, L.
Since GPS, as a commonly used positioning system in outdoor environments, cannot be used in indoor environments, localization methods suitable for indoor environments are still being investigated. The fingerprinting method stands out from other indoor localization methods because it can use existing signal sources and can be implemented by ubiquitous devices such as mobile phones. In this study, several classification algorithms used in the fingerprinting method are applied to two datasets obtained from two different environments (home and workplace). Among these classification algorithms, Random Forest achieved the best results with 87% and 74% accuracy rates for these datasets. These results are close to the results reported in previous studies, and the accuracy of the algorithms varies depending on the environment in which the dataset has been formed. © 2015 IEEE.
Citation - Scopus: 3
Personal Mark Density-Based High-Performance Optical Mark Recognition (OMR) System Using K-Means Clustering Algorithm
(Springer, 2024) Sancar, Y.; Yavuz, U.; Karabey Aksakalli, I.
To evaluate multiple choice question tests, optical forms are commonly used for large-scale exams and these forms are read by the OMR (Optical Mark Recognition) scanners. However, OMR scanners often misinterpret marks that have not been fully erased, which can lead to incorrect readings. To overcome that shortcoming and reduce the time and labor lost in the assessment process, we developed a novel system based on the density of each individual’s markings, providing a more personalized and accurate approach. Instead of reading according to a specific optical form template, a dynamic and flexible structure was generated where users can create own templates and obtain the model that reads according to that template. We also optimized certain aspects of the system for efficiency, such as image memory transfer and QR code reading. These optimizations significantly increase the performance of the OMR scanners. One of the key issues addressed is inaccurate reading of OMR scanners when a student doesn’t fully erase their markings or when markings are faint. After the scanning process, the proposed approach uses a K-means clustering algorithm to classify different density markings. This technique identifies each student’s personal marking density, enabling a more accurate interpretation of their responses. According to the experimental results, we performed 97.7% improvement compared to the misread optics scanned by the conventional OMR devices. In tests performed on 265.816 optical forms, we obtained an accuracy rate of 99.98% and a reading time of 0.12 seconds per optical form. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2024.
Citation - Scopus: 4
Text Classification-Based Petition Recognition and Routing System: A Turkish Case Study
(Springer Science and Business Media B.V., 2023) Sancar, Y.; Karabey Aksakalli, I.; Karacali, T.
Large-scale enterprises include many units and various electronic petitions are manually submitted to the relevant units by authorized staff. However, this delivery process is time-consuming in large-scale institutions and the petitions may not be delivered to the relevant units on time. The manual categorization of the petitions and the time wasted while transmitting the responses given to the recipients in the same way leads to delay in business life. That administrators cannot follow the petitions shows a need for a petition recognition system that automatically directs the relevant unit according to its content. In this study, electronic petitions sent from any unit of the institution are processed through a petition recognition and routing system. The system offers a solution to direct petitions each relevant unit according to their subjects. In this system, a printed document is scanned through the OCR (Optical Character Recognition) techniques and the characters are extracted from the digital petitions. After the pre-processing and feature extraction phase, the petitions are categorized using various machine learning classification methods, and the proposed routing system automatically detects the most successful classification method to direct the petitions to the relevant units. The experimental results show that the proposed petition recognition and routing system can classify the petitions by 0.951 accuracy rate and 0.94 f-macro value using Stochastic Gradient Descent classifier with BoWtfidf vectorized method. The performance of the proposed petition classification and routing system is reasonable for the end users. Based on our investigation, this study is the first in its area that contributes a novel petition benchmark dataset and addresses the petition classification issue by combining OCR, natural language processes, and machine learning techniques. The novel petition dataset is thought to pave the way for further research in petition text classification. © 2023, The Author(s), under exclusive licence to Bharati Vidyapeeth's Institute of Computer Applications and Management.
Citation - Scopus: 2
Utilization of Room-To Transition Time in Wi-Fi Fingerprint-Based Indoor Localization
(Institute of Electrical and Electronics Engineers Inc., 2015) Karabey Aksakalli, I.; Bayindir, L.
In indoor localization applications, many different methods have been proposed to increase positioning accuracy. Among these methods, fingerprint-based techniques are generally preferred because they use existing resources such as Wi-Fi, Bluetooth, FM signals, etc., and can be implemented on commonly used devices such as mobile phones. In this paper, we evaluate different Wi-Fi fingerprint-based methods on two datasets (with and without room-to-room transition features) created from the same environment, and we investigate the impact of room-to-room transition features on classification performance. To the best of our knowledge, transition time between rooms has not been used in past studies on fingerprint-based indoor localization. This information is of significant importance, due to the physical distance between rooms. Therefore, in this study source room and transition time to a target room have been included as features in addition to signal sources and signal strength values in the target room. From preliminary experimental results we observed that the transition time between rooms increases the performance of all tested positioning algorithms, with the Back-propagation classifier showing the best performance increase (13%). © 2015 IEEE.