Scopus İndeksli Yayınlar Koleksiyonu
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12573/395
Browse
6 results
Search Results
Conference Object Citation - WoS: 3Citation - Scopus: 4Fine Tuning DeepSeek and Llama Large Language Models with LoRA(IEEE, 2025-06-25) Uluirmak, Bugra Alperen; Kurban, RifatIn this paper, Low-Rank Adaptation (LoRA) finetuning of two different large language models (DeepSeek R1 Distill 8B and Llama3.1 8B) was performed using the Turkish dataset. Training was performed on Google Colab using A100 40 GB GPU, while the testing phase was carried out on Runpod using L4 24 GB GPU. The 64.6 thousand row dataset was transformed into question-answer pairs from the fields of agriculture, education, law and sustainability. In the testing phase, 40 test questions were asked for each model via Ollama web UI and the results were supported with graphs and detailed tables. It was observed that the performance of the existing language models improved with the fine-tuning method.Conference Object Citation - Scopus: 4Re-Exploring the Kayseri Culture Route by Using Deep Learning for Cultural Heritage Image Classification Cultural Heritage Image Classification by Using Deep Learning: Kayseri Culture Route(Association for Computing Machinery, 2024-05-25) Kevseroğlu, Ozlem; Kurban, RifatThe categorization of images captured during the documentation of architectural structures is a crucial aspect of preserving cultural heritage in digital form. Dealing with a large volume of images makes this categorization process laborious and time-consuming, often leading to errors. Introducing automatic techniques to aid in sorting would streamline this process, enhancing the efficiency of digital documentation. Proper classification of these images facilitates improved organization and more effective searches using specific terms, thereby aiding in the analysis and interpretation of the heritage asset. This study primarily focuses on applying deep learning techniques, specifically SqueezeNet convolutional neural networks (CNNs), for classifying images of architectural heritage. The effectiveness of training these networks from scratch versus fine-tuning pre-existing models is examined. In this study, we concentrate on identifying significant elements within images of buildings with architectural heritage significance of Kayseri Culture Route. Since no suitable datasets for network training were found, a new dataset was created. Transfer learning enables the use of pre-trained convolutional neural networks to specific image classification tasks. In the experiments, 99.8% of classification accuracy have been achieved by using SqueezeNet, suggesting that the implementation of the technique can substantially enhance the digital documentation of architectural heritage. © 2024 Elsevier B.V., All rights reserved.Article Citation - WoS: 3Citation - Scopus: 8Multi-Focus Image Fusion by Using Swarm and Physics Based Metaheuristic Algorithms: A Comparative Study With Archimedes, Atomic Orbital Search, Equilibrium, Particle Swarm, Artificial Bee Colony and Jellyfish Search Optimizers(Springer, 2023-09-07) Cakiroglu, Fatma; Kurban, Rifat; Durmus, Ali; Karakose, ErcanThe lenses focus only on the objects at a specific distance when an image is captured, the objects at other distances look blurred. This is referred to as the limited depth of field problem, and several attempts exist to solve this problem. Multi-focus image fusion is one of the most used methods when solving this problem. A clear image of the whole scene is obtained by fusing at least two different images obtained with different focuses. Block-based methods are one of the most used methods for multi-focus fusion at the pixel-level. The size of the block to be used is an important factor for determining the performance of the fusion. Thus, the block size must be optimized. In this study, the comparison between the swarm-based and physics-based algorithms is made to determine the optimal block size. The comparison has been made among the following optimization methods which are, namely, Archimedes Optimization Algorithm (AOA), Atomic Orbital Search (AOS) and Equilibrium Optimizer (EO) from the physics-based algorithms and Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC) and Jellyfish Search Algorithm (JSA) from swarm-based algorithms. The swarm-based ABC and JSA algorithms have shown a better performance when compared to physics-based methods. Moreover, meta-heuristic algorithms, in general, are more adaptive compared to the traditional fusion methods.Article Citation - WoS: 20Citation - Scopus: 32Classification of Apple Images Using Support Vector Machines and Deep Residual Networks(Springer London Ltd, 2023-02-21) Adige, Sevim; Kurban, Rifat; Durmus, Ali; Karakose, ErcanOne of the most important problems for farmers who produce large amounts of apples is the classification of the apples according to their types in a short time without handling them. Support vector machines (SVM) and deep residual networks (ResNet-50) are machine learning methods that are able to solve general classification situations. In this study, the classification of apple varieties according to their genus is made using machine learning algorithms. A database is created by capturing 120 images from six different apple species. Bag of visual words (BoVW) treat image features as words representing a sparse vector of occurrences over the vocabulary. BoVW features are classified using SVM. On the other hand, ResNet-50 is a convolutional neural network that is 50 layers deep with embedded feature extraction layers. The pre-trained ResNet-50 architecture is retrained for apple classification using transfer learning. In the experiments, our dataset is divided into three cases: Case 1: 40% train, 60% test; Case 2: 60% train, 40% test; and Case 3: 80% train, 20% test. As a result, the linear, Gaussian, and polynomial kernel functions used in the BoVW + SVM algorithm achieved 88%, 92%, and 96% accuracy in Case 3, respectively. In the ResNet-50 classification, the root-mean-square propagation (rmsprop), adaptive moment estimation (adam), and stochastic gradient descent with momentum (sgdm) training algorithms achieved 86%, 89%, and 90% accuracy, respectively, in the set of Case 3.Conference Object Benchmarking CNN Architectures for Eye Disease Detection With Transfer Learning Techniques(Institute of Electrical and Electronics Engineers Inc., 2025-06-27) Keles, Tolgahan; Aykanat, Muhammet Ali; Kurban, RifatIn this study, convolutional neural networks (CNN)-based approaches were compared to classify eye diseases using transfer learning techniques. A series of data augmentation strategies, including random rotation, shifting, shearing, zooming, and horizontal flipping, were applied to increase the training data's robustness and diversity. Several state-of-the-art CNNs, including ResNet50, VGG19, EfficientNetB0, Xception, InceptionV3, DenseNet121, MobileNetV2, NASNetMobile, and ConvNeXtBase, were fine-tuned through transfer learning. During training, models were evaluated based on their accuracy, training time, and validation performance, while early stopping mechanisms were employed to prevent overfitting. Experimental results demonstrated that DenseNet121 achieved the highest validation accuracy (72%) during the training phase and the best test set performance with an accuracy of 68% and an AUC-ROC of 0.93. MobileNetV2, on the other hand, provided a strong balance between classification accuracy (65%) and low inference time (7.28 ms), making it appropriate for real-time uses. The findings highlight the importance of selecting appropriate architectures by considering both predictive performance and computational efficiency, particularly in the context of medical imaging, where real-world deployment constraints are critical. © 2025 Elsevier B.V., All rights reserved.Conference Object Citation - Scopus: 1A Comprehensive Investigation into Strip Steel Defect Detection Using Traditional Machine Learning and Deep Learning Models(IEEE, 2025-05-23) Erkantarci, Betul; Kurban, Rifat; Bakal, Mehmet Gokhan; Kose, AbdulkadirThe steel manufacturing sector places great importance on guaranteeing the quality of strip steel products, which has led to a thorough investigation of defect detection approaches. This work conducts a comparative analysis of traditional machine learning and deep learning models to determine their efficacy in detecting defects in strip steel. Our analysis is based on a dataset that includes a variety of images of strip steel surfaces showing different types of defects. In this work, we adopt image preprocessing techniques to improve the quality of input images prior to the application of classification methods. We employ traditional ML algorithms including Support Vector Machine and Random Forest, and deep learning model AlexNet Convolutional Neural Networks for effective defect classification. Consequently, we present comparative evaluations that highlight the strengths and weaknesses of each approach, considering accuracy scores.