Browsing by Author "Ozhan, Davut"

Now showing 1 - 4 of 4

Citation - WoS: 8
Citation - Scopus: 21
Automatic Detection of Brain Tumors With the Aid of Ensemble Deep Learning Architectures and Class Activation Map Indicators by Employing Magnetic Resonance Images
(Elsevier, 2024) Turk, Omer; Ozhan, Davut; Acar, Emrullah; Akinci, Tahir Cetin; Yilmaz, Musa
Today, as in every life-threatening disease, early diagnosis of brain tumors plays a life-saving role. The brain tumor is formed by the transformation of brain cells from their normal structures into abnormal cell structures. These formed abnormal cells begin to form in masses in the brain regions. Nowadays, many different techniques are employed to detect these tumor masses, and the most common of these techniques is Magnetic Resonance Imaging (MRI). In this study, it is aimed to automatically detect brain tumors with the help of ensemble deep learning architectures (ResNet50, VGG19, InceptionV3 and MobileNet) and Class Activation Maps (CAMs) indicators by employing MRI images. The proposed system was implemented in three stages. In the first stage, it was determined whether there was a tumor in the MR images Tumor) were detected from MR images (Multi-class Approach). In the last stage, CAMs of each tumor group were created as an alternative tool to facilitate the work of specialists in tumor detection. The results showed that the overall accuracy of the binary approach was calculated as 100% on the ResNet50, InceptionV3 and MobileNet architectures, and 99.71% on the VGG19 architecture. Moreover, the accuracy values of 96.45% with ResNet50, 93.40% with VGG19, 85.03% with InceptionV3 and 89.34% with MobileNet architectures were obtained in the multi-class approach.
Citation - WoS: 33
Citation - Scopus: 39
Design and Implementation of a New Contactless Triple Piezoelectrics Wind Energy Harvester
(Pergamon-Elsevier Science Ltd, 2017) Kurt, Erol; Cottone, Francesco; Uzun, Yunus; Orfei, Francesco; Mattarelli, Maurizio; Ozhan, Davut
The features of the new designed and constructed harvester are examined. The harvested power of three piezoelectric layers having different masses (i.e. different natural frequencies) has been explored. These layers have the same length around the harvester body, whereas a permanent magnet (PM) attached to the shaft rotates by low speed wind and this PM repels these three piezoelectric layers with a 120 phase shift. Since PM and the PMs located to the tip of the layers do not contact, this system improves the lifetime of the harvester. The measured harvested power in the low wind speeds (i.e. 1.75 m/s) is of the order of 0.2 mu W. The waveform includes many subharmonic and superharmonic components, hence the total harmonic distortion (THD) is found around 130%, which is fairly high due to nonlinear effects. Although the system shows an high THD, the 20% of the signal can be rectified and stored in the capacitor for the use of harvested energy. A scenario has also been created for a resistive load of R-L, =1 M Omega and 100 k Omega for various wind speeds and it has been proven that the harvester can feed the load at even lower wind speeds. In addition, extra power beyond the usage of the load can be stored into the capacitor. The proposed harvester and its rectifying unit can be a good solution for the energy conversion procedures of low-power required machines. (C) 2017 Hydrogen Energy Publications LLC. Published by Elsevier Ltd. All rights reserved.
An Energy Storage Unit Design for a Piezoelectric Wind Energy Harvester with a High Total Harmonic Distortion
(MDPI, 2025) Ozhan, Davut; Kurt, Erol
A new energy storage unit, which is fed by a piezoelectric wind energy harvester, is explored. The outputs of a three-phase piezoelectric wind energy device have been initially recorded from the laboratory experiments. Following the records of voltage outputs, the power ranges of the device were measured at several hundred microwatts. The main issue of piezoelectric voltage generation is that voltage waveforms of piezoelectric materials have high total harmonic distortion (THD) with incredibly high subharmonics and superharmonics. Therefore, such a material reply causes a certain power loss at the output of the wind energy generator. In order to fix this problem, we propose a combination of a rectifier and a storage system, where they can operate compatibly under high THD rates (i.e., 125%). Due to high THD values, current-voltage characteristics are not linear-dependent; indeed, because of capacitive effect of the piezoelectric (i.e., lead zirconium titanite) material, harvested power from the material is reduced by nearly a factor of 20% in the output. That also negatively affects the storage on the Li-based battery. In order to compensate, the output waveform of the device, the waveforms, which are received from the energy-harvester device, are first rectified by a full-wave rectifier that has a maximum power point tracking (MPPT) unit. The SOC values prove that almost 40% of the charge is stored in 1.2 s under moderate wind speeds, such as 6.1 m/s. To conclude, a better harvesting performance has been obtained by storing the energy into the Li-ion battery under a current-voltage-controlled boost converter technique.
Towards Social Media Analytics and Real-Time Trolls Identification Automation Using Artificial Intelligence
(IEEE, 2025) Raghunathan, Keerthikant; Syed, Dabeeruddin; Zainab, Ameema; Vuppala, Mounika; Ozhan, Davut; Refaat, Shady S.
Social media platforms have become a common ground to induce political discussions and manipulating public opinion. This has led to the inception of trolling that involves diverting public opinion from facts, driving people into emotionally charged and appealing discussions, and spreading disinformation against an organization, entity or a country for political or other gains. Trolling is conducted by humans or programmed bots. In the recent past, there have been numerous manipulative and malicious campaigns to spread fake news about different nations and economies on social media. Rather than focusing on individual accounts, it is crucial to discover manipulative campaigns. This involves challenges such as recognizing complex patterns, computational complexity, and real-time performance. Our proposed solution of real-time trolls identification automation is based on social media analytics using deep machine learning. Real-time ability is achieved using text stream clustering, and the design approach is evaluated on real-world tweets. The current performed work utilizes fine tuning large language models to apprehend a higher degree of complexity and employs a distilled form of Bidirectional Encoder Representations from Transformers models to obtain high accuracy of detection. A correspondence analysis is beneficial to map noun-verb relationships in structured data. With the proposed approach, an accuracy of 92% was achieved.