Araştırma Çıktıları | WoS | Scopus | TR-Dizin | PubMed
Permanent URI for this communityhttps://hdl.handle.net/20.500.14719/1741
Browse
10 results
Search Results
Publication Metadata only Consumer Preference Estimation Using EEG Signals and Deep Learning(IEEE, 2024) Ceylan, Burak; Cekic, Yalcin; Akan, Aydin; Bahcesehir University; Izmir Ekonomi UniversitesiEmotion estimation is an extremely critical and current research topic for human-computer interaction. In this study, a liking estimation method using electroencephalogram (EEG) signals is proposed to be used in neuromarketing studies. EEG data recorded while participants watch the advertisement videos of two different automobile brands are processed with deep learning techniques to estimate their liking status. After watching the videos, participants were presented with selected image sections from the advertisements (front view, console, side view, rear view, stop lamp, brand logo and front grille) and were asked to rate their liking by scoring from 1 to 5. EEG signals corresponding to these regions were converted into a two dimensional and RGB colored image using the short-time Fourier transform (STFT) method, and liking status was estimated using Deep Learning. The successful results obtained show that the proposed method can be used in neuromarketing studies.Publication Metadata only Creation of Annotated Synthetic UAV Video Dataset for Object Detection and Tracking(IEEE, 2023) Yilmaz, Can; Maras, Bahri; Arica, Nafiz; Ertuzun, Aysin Baytan; Bahcesehir University; Bogazici University; Piri Reis UniversityIn order for object detection and tracking in videos obtained from unmanned aerial vehicles (UAVs) by deep convolutional neural networks (DCNN), extensive ground truth optical flow, occlusion and segmentation datasets, of various objects or vehicles, are required during the training and testing processes. The mentioned ground truth informations are not widely available in the literature due to the difficulty of labeling or extracting them from real-life recorded UAV video images. In this study, ground truth optical flow, occlusion and segmentation datasets were produced synthetically for the first time with the UAV point of view in a novel way, so as to fill the gap in literature. The ground truth datasets were created for each vehicle by subjecting the triangles (mesh) automatically generated by the Unity engine to the homography method. With this method, 1920x1080 and 250x250 sized synthetic datasets consisting of 100 scenarios were obtained.Publication Metadata only Neural Architecture Search Using Differential Evolution in MAML Framework for Few-Shot Classification Problems(SPRINGER INTERNATIONAL PUBLISHING AG, 2023) Gulcu, Ayla; Kus, Zeki; DiGaspero, L; Festa, P; Nakib, A; Pavone, M; Bahcesehir UniversityModel-Agnostic Meta-Learning (MAML) algorithm is an optimization based meta-learning algorithm which aims to find a good initial state of the neural network that can then be adapted to any novel task using a few optimization steps. In this study, we take MAML with a simple four-block convolution architecture as our baseline, and try to improve its few-shot classification performance by using an architecture generated automatically through the neural architecture search process. We use differential evolution algorithm as the search strategy for searching over cells within a predefined search space. We have performed our experiments using two well-known few-shot classification datasets, mini-ImageNet and FC100 dataset. For each of those datasets, the performance of the original MAML is compared to the performance of our MAML-NAS model under both 1-shot 5-way and 5-shot 5-way settings. The results reveal that MAML-NAS results in better or at least comparable accuracy values for both of the datasets in all settings. More importantly, this performance is achieved by much simpler architectures, that is architectures requiring less floating-point operations.Publication Metadata only Enhancing Object Detection Algorithms by Synthetic Aerial Images(IEEE, 2023) Yilmaz, Can; Maras, Bahri; Yilmaz, Gorkem; Ceylan, Goksu; Hamamcioglu, Onder; Arica, Nafiz; Ertuzun, Aysin Baytan; Bahcesehir University; Bogazici University; Bahcesehir University; Piri Reis UniversityIn order to accurately perform object detection by deep convolutional neural networks (DCNN) in videos, obtained from unmanned aerial vehicles (UAVs), many example images of objects containing annotations such as ground truth class information, bounding box, optical flow, occlusion and segmentation are required. Due to the difficulties faced during annotation of scenarios, and due to the inadequacy of the scenario diversity resulting from environmental conditions, a dataset containing above mentioned ground truths has not been found in the literature. In this study, synthetic aerial images with various annotation information were created in different scenarios while composing virtual worlds, and enhancing object detection algorithms is aimed. Enhancement of detection results of DCNN based object detection algorithms, trained with the support of synthetic aerial images, on real-world aerial images significantly, was observed during the experiments, conducted.Publication Metadata only Comparison of Six Different Dielectric Substrates in Microstrip Patch Antenna Design(IEEE, 2024) Goksel, Fatih; Karamzadeh, Saeid; Bahcesehir University; Bahcesehir UniversityMicrostrip patch antenna (MSPA) is the most popular antenna type due to the small size, easy to fabricate, simple and inexpensive construction, lightweight and low volume, conformity with planar and non-planar surfaces, dual and triple frequency operation in recent years. Especially, due to the becoming the size of the devices becoming smaller, desire to design MSPA has gained more importance than before. In this study, as a result of comparison between six different types of antennas, showed us which of the substrate or substrates give us the best performance of the newly designed patch antenna could be chosen. Based on this, in this study, designed rectangular MSPA simulated with six different substrates which are FR4 Glass Epoxy with 4.3, ROGERS DT5880 with 2.2, ROGERS 4003C with 3.3, Teflon with 2.1, Arlon AD300A with 3 and last Alumina with 9.9 dielectric constant. The study includes a copper patch antenna and copper ground working at 5.8 GHz. The six substrates are kept constant at 1.6 mm height. Good results gained at the desired frequency of 5.8 GHz antenna which is at the resonant frequency between 5.7-5.9 GHz. The main aim of this study is to support their research and create an idea about the six different substrates which can be prepared by researchers while choosing the substrate of the MSPAs.Publication Metadata only Graph Neural Networks Based Approach for Interpersonal Relationship Classification in Images(IEEE, 2023) Akay, Simge; Arica, Nafiz; Bahcesehir University; Piri Reis UniversityInterpersonal relationships, which are an integral part of our social life, demonstrate how people connect and interact within society. Describing the relationship between two individuals in images depends on many different factors and attributes. In this study, a Graph Neural Network (GNN) based approach is proposed that focuses on the important attributes for describing the relationship between two individuals in images. In the proposed method, each attribute that will be used to describe the relationship is defined as a GNN node. Then, the meaningless connections between nodes are pruned with a pruning operation to obtain the ideal GNN model. In this study, a more robust GNN for classifying interpersonal relationships is created by using rich attributes, unlike the literature, and obtaining significant connections between nodes through pruning. The experiments conducted in this study showed that both using a wider GNN and pruning operation improve the classification performance.Publication Metadata only Anomaly Detection in Hyperspectral Images with High Dimensional Model Representation(IEEE, 2024) Bodur, Derya; Ozay, Evrim Korkmaz; Tunga, Burcu; Bahcesehir University; Istanbul Technical UniversityHyperspectral imaging (HSI) is an important imaging technology that enables the high sensitivity and accuracy analysis of wide areas through its multispectral band structure. Anomaly detection in hyperspectral images can be defined as the detection of pixels that do not belong to the background and whose spectral properties are unknown. In this study, a method based on High Dimensional Model Representation (HDMR) is proposed for detecting anomalies in hyperspectral images. The effectiveness of the proposed method has been tested on a sample image and compared with commonly used anomaly detection methods. The HDMR-based anomaly detection algorithm has shown to be more effective in suppressing the background compared to other methods by making anomaly pixels more visible.Publication Metadata only Detection of Human Respiratory and Range by IRUWB Radar(IEEE, 2023) Shakfa, Ali; Esmaeilishahir, Mahdi; Karamzadeh, Saeid; Bahcesehir UniversityThe technology of Impulse Radio Ultra-Wide-Band (IR-UWB) radar is very promising for non-contact sensing applications, such as detecting human vital signs like heart rate (HR) and respiration rate (RR), as well as determining the location of a person. This research paper introduces a technique for detecting the respiration rate and location of a stationary human. The method involves using preprocessing techniques to improve clutter reduction and estimating the frequency domain of the signals. Low pass and bandpass filters as preprocessing methods are presented. The Fast Fourier Transform (FFT) processing method is used to determine the human RR in the frequency domain. The proposed approach is evaluated using a P440 (IR-UWB) radar and shows promising results. The experiments demonstrate that the proposed method is effective in accurately detecting the vital signs of a stationary single or multiple persons.Publication Metadata only Insult Detection in the Turkish Language Through Different Machine Learning Algorithms(IEEE, 2023) Ozgen, Kerem; Rada, Lavdie; Bahcesehir UniversityIn this research paper, we propose to use the Turkish Court of Cassation- Yargitay- cases to build a dataset for insult detection tasks and compare machine learning models trained on this dataset. We accumulated studies available in the literature compiling Court of Cassation cases and generated a train and test set for testing machine learning algorithms for insult detection. Although machine learning is not capable of understanding the legal context, cultural background, and the nature of insults or non-insults, it can help identify insults with proper training data created by experts. As far as for the authors knowledge this is the first study to use machine learning for the purpose of automatically distinguishing between insult and non-insult cases within the Turkish justice system. Our research, though its is in its first steps, represents a significant contribution to the field, as it addresses a gap in the existing literature and provides a machine learning approach to improving the efficiency and accuracy of legal decision-making.Publication Metadata only Detecting Anomalies in Information Assets with Graph Signal Processing(IEEE, 2024) Akgun, Ayhan; Bahcesehir UniversityThe significant business requirement for organizations utilizing technology-based services is the production of appropriate services and applications according to business needs, and the sustainable operation of the infrastructures providing these services. Even as organizations increasingly procure services externally due to new technological advancements, there remains a need to monitor Information Technology structures, configuration, and version requirements. The Information Technology Infrastructure Library (ITIL) and its subset, the Configuration Management Database (CMDB), provide a roadmap for these approaches. However, as technologies advance, this management need can become even more complex. Recently, with the increasing use of graph-based perspectives, both in artificial intelligence studies and in relational complex systems such as social networks and network structures, managing definitions in multidimensional space systems, such as objects in knowledge management systems using graph algorithms, has become easier. Within these studies, well-known artificial intelligence tasks such as prediction, classification, or anomaly detection can be more controllable.