Makale Koleksiyonu

Bu koleksiyon için kalıcı URI

https://hdl.handle.net/20.500.12587/8637

Listeleniyor 1 - 20 / 66

Application of deep ensemble learning for palm disease detection in smart agriculture
(Cell Press, 2024) Savaş, Serkan
Agriculture has notably become one of the fields experiencing intensive digital transformation. Leveraging state-of-the-art techniques in this domain has provided numerous advantages for agricultural activities. Deep learning (DL) algorithms have proven beneficial in addressing various agricultural challenges. This study presents a comprehensive investigation into applying DL models for palm disease detection and classification in the context of smart agriculture. The research aims to address the limitations observed in previous studies and improve the robustness and generalizability of the results. To achieve this, a two-stage optimization methodology is employed. First, transfer learning and fine-tuning techniques are applied using various pretrained deep neural network models. The experiments show promising results, with all models achieving high accuracy rates during training and validation. Furthermore, their performance on unseen test data is also assessed to ensure practical applicability. The top-performing models are MobileNetV2 (92.48 %), ResNet (92.42 %), ResNetRS50 (92.30 %), and DenseNet121 (92.01 %). Second, a deep ensemble learning approach is applied to enhance the models' generalization capability further. The best-performing models with different criteria are combined using the ensemble technique, resulting in remarkable improvements in disease detection tasks. DELM1 emerges as the most successful ensemble model, achieving an ROC AUC Score of 99%. This study demonstrates the effectiveness of deep ensemble learning models in palm disease detection and classification for smart agriculture applications. The findings contribute to advancing disease detection systems and emphasize the potential of ensemble learning. The study provides valuable insights for future research, guiding the application of DL techniques to address critical agricultural challenges and improve crop health monitoring systems. Another contribution is combining various plant diseases and insect pest classes using diverse datasets. A comprehensive classification system is achieved by considering different disease classes and stages within the white scale category, improving the model's robustness.
A novel image watermarking scheme using ULV decomposition
(Elsevier Gmbh, 2022) Horasan, Fahrettin
Matrix decompositions play an important role in most of watermarking techniques. Especially Singular Value Decomposition (SVD) is one of the most preferred techniques. In addition, matrix decompositions such as Non-negative Matrix Factorization (NMF), QR, LU are some of the methods used in previous studies. In this study, a new scheme using ULV Decomposition (ULVD) technique is proposed as an alternative matrix decomposition. In this scheme, which is a frequency-based technique using R-level Discrete Wavelet Transform (DWT), the scaling factor is determined adaptively according to the cover image and watermark used. Another issue to be solved in watermarking studies is the False Positive Problem (FFP). For this purpose, a control mechanism is used against the FFP in watermark embedding and watermark extracting processes. In addition, the watermarking process is carried out in any size, depending on the size of the cover image. As a result, the performed experiments show that the proposed scheme provides high imperceptibility and robustness.
Determining the Location of Tibial Fracture of Dog and Cat Using Hybridized Mask R-CNN Architecture
(Kafkas Univ, Veteriner Fakultesi Dergisi, 2021) Baydan, Berker; Barışçı, Necaattin; Ünver, Halil Murat
The aim of this study is to hybridize the original backbone structure used in the Mask R-CNN framework, and to detect fracture location in dog and cat tibia fractures faster and with higher performance. With the hybrid study, it will be ensured that veterinarians help diagnose fractures on the tibia with higher accuracy by using a computerized system. In this study, a total of 518 dog and cat fracture tibia images that obtained from universities and institutions were used. F1 score value of this study on total dataset was found to be 85.8%. F1 score value of this study on dog dataset was found to be 87.8%. F1 score value of this study on cat dataset was found to be 77.7%. With the developed hybrid system, it was determined that the localization of the fracture in an average tibia image took 2.88 seconds. The results of the study showed that the hybrid system developed would be beneficial in terms of protecting animal health by making more successful and faster detections than the original Mask R-CNN architecture.
Detection of tibial fractures in cats and dogs with deep learning
(Ankara Univ Press, 2021) Baydan, Berker; Ünver, Halil Murat
The aim of this study is to classify tibia (fracture/no fracture) on whole/partial body digital images of cats and dogs, and to localize the fracture on fracture tibia by using deep learning methods. This study provides to diagnose fracture on tibia more accurately, quickly and safe for clinicians. In this study, a total of 1488 dog and cat images that were obtained from universities and institutions were used. Three different studies were implemented to detect fracture tibia. In the first phase of the first study, tibia was classified automatically as fracture or no fracture with Mask R-CNN. In the second phase, the fracture location in the fracture tibia image that obtained from the first phase was localized with Mask R-CNN. In the second study, the fracture location was directly localized with Mask R-CNN. In the third study, fracture location in the fracture tibia that obtained from the first phase of first study was localized with SSD. The accuracy and F1 score values in first phase of first study were 74% and 85%, respectively and F1 score value in second phase of first study was 84.5%. The accuracy and F1 score of second study were 52.1% and 68.5%, respectively. The F1 score of third study was 46.2%. The results of the research showed that the first study was promising for detection of fractures in the tibia and the dissemination of the fracture diagnosis with the help of such smart systems would also be beneficial for animal welfare.
Keyword Extraction for Search Engine Optimization Using Latent Semantic Analysis
(Gazi Univ, 2021) Horasan, Fahrettin
It is now difficult to access desired information in the Internet world. Search engines are always trying to overcome this difficulty. However, web pages that cannot reach their target audience in search engines cannot become popular. For this reason, search engine optimization is done to increase the visibility in search engines. In this process, a few keywords are selected from the textual content added to the web page. A responsible person who is knowledgeable about the content and search engine optimization is required to determine these words. Otherwise, an effective optimization study cannot be obtained. In this study, the keyword extraction from textual data with latent semantic analysis technique was performed. The latent semantic analysis technique models the relations between documents/sentences and terms in the text using linear algebra. According to the similarity values of the terms in the resulting vector space, the words that best represent the text are listed. This allows people without knowledge of the SEO process and content to add content that complies with the SEO criteria. Thus, with this method, both financial expenses are reduced and the opportunity to reach the target audience of web pages is provided.
U2-Net Segmentation And Multi-Label Cnn Classification Of Wheat Varieties
(Konya Teknik Univ, 2024) Argun, Mustafa Şamil; Türk, Fuat; Civelek, Zafer
There are many varieties of wheat grown around the world. In addition, they have different physiological states such as vitreous and yellow berry. These reasons make it difficult to classify wheat by experts. In this study, a workflow was carried out for both segmentation of wheat according to its vitreous/yellow berry grain status and classification according to variety. Unlike previous studies, automatic segmentation of wheat images was carried out with the U2-NET architecture. Thus, roughness and shadows on the image are minimized. This increased the level of success in classification. The newly proposed CNN architecture is run in two stages. In the first stage, wheat was sorted as vitreous-yellow berry. In the second stage, these separated wheats were grouped by multi-label classification. Experimental results showed that the accuracy for binary classification was 98.71% and the multi-label classification average accuracy was 89.5%. The results showed that the proposed study has the potential to contribute to making the wheat classification process more reliable, effective, and objective by helping the experts.
LSTM Network Based Sentiment Analysis for Customer Reviews
(Gazi Univ, 2022) Bilen, Burhan; Horasan, Fahrettin
Continuously increasing data bring new problems and problems usually reveal new research areas. One of the new areas is Sentiment Analysis. This field has some difficulties. The fact that people have complex sentiments is the main cause of the difficulty, but this has not prevented the progress of the studies in this field. Sentiment analysis is generally used to obtain information about persons by collecting their texts or expressions. Sentiment analysis can sometimes bring serious benefits. In this study, with singular tag-plural class approach, a binary classification was performed. An LSTM network and several machine learning models were tested. The dataset collected in Turkish, and Stanford Large Movie Reviews datasets were used in this study. Due to the noise in the dataset, the Zemberek NLP Library for Turkic Languages and Regular Expression techniques were used to normalize and clean texts, later, the data were transformed into vector sequences. The preprocessing process made 2% increase to the model performance on the Turkish Customer Reviews dataset. The model was established using an LSTM network. Our model showed better performance than Machine Learning techniques and achieved an accuracy of 90.59% on the Turkish dataset and an accuracy of 89.02% on the IMDB dataset.
A Linear Time Pattern Based Algorithm for N-Queens Problem
(Gazi Univ, 2022) Karabulut, Bergen; Erguzen, Atilla; Unver, Halil Murat
The n-queens problem is the placing of n number of queens on an nxn chessboard so that no two queens attack each other. This problem is important due to various usage fields such as VLSI testing, traffic control job scheduling, data routing, dead-lock or blockage prevention, digital image processing and parallel memory storage schemes mentioned in the literature. Besides, this problem has been used as a benchmark for developing new artificial intelligence search techniques. However, it is known that backtracking algorithms, one of the most frequently used algorithms to solve this problem, cannot produce all solutions in large n values due to the exponentially growing time complexity. Therefore, various methods have been proposed for producing one or more solutions, not all solutions for each n value. In this study, a pattern based approach that produces at least one unique solution for all n values (n>3) was detected. By using this pattern, a new algorithm that produces at least one unique solution for even very large n values in linear time was developed. The developed algorithm with.(n) time complexity produces quite faster solution to n-queens problem and even in some values, this algorithm produces (n-1)/2 unique solutions in linear time.
Design of a Tracking Welding Robot Automation System for Manufacturing of Steam and Heating Boilers
(2018) Ersöz, Süleyman; Türker, Ahmet Kürşad; Aktepe, Adnan; Atabaş, İrfan; Kokoç, Melda
For satisfying customers companies want to respond to customer requests on time. At the same time, they expect production process to be completed with low cost and low loss. For this reason, the importance of mechanization and automation in production sector has increased. As a result, companies have begun to give more importance to robotic systems, which are the basic components of automation systems. Despite the likelihood of mistakes caused by physiological and mental states of humans, these systems can perform operations precisely without any variability. In this study, an application was carried out for the automation of welding process of industrial type boilers in different sizes and features. For products of which standard measurements or welding operations are difficult to perform manually, a robotic system was proposed in which measurement and welding operations can be performed automatically. In addition, operators are prevented from exposure to gas and light via the proposed system which enables a safer working condition.
Motion Capture Technology in Sports Science: 3D Virtual Sports Platform with Kinect
(2017) Erbay, H; Kutlu, M
Düzenli spor yapmanın yaşam ve sağlık kalitesi için yararlı olduğu kabul edilmektedir. Fakat düzenli spor yapma olanaklarının özellikle gelişmekte olan ülkelerde yeterince yaygınlaşmadığı gözlenmektedir. İnternet ve Kinect gibi güncel ve gelişmiş teknolojilerin sağladığı olanaklar ile, spor yapamayanlara fiziksel olarak aynı ortamda bulunmaksızın, sanal eğitmen kontrolünde spor yapabilme olanağı sunmanın alternatif bir çözüm yaratabileceği düşünülmektedir. Bu doğrultudaki çalışmalarla, Kinect üç boyutlu sanal spor platformu tasarlanıp geliştirilmiştir. Geliştirilen platformun özellikleri, yeni teknolojilerin avantajlarının kullanıldığı platformlar ile alternatif spor yapma olanakları gözden geçirilmektedir
Bulut Tabanlı Öğrenme Yönetim Sistemi Seçiminde Bulanık Çok Kriterli Karar Analizi Yaklaşımı
(2020) Özcan, Hakan; Emiroğlu, Bülent Gürsel
Bulut bilişim teknolojisinin gelişmesiyle öğrenme yönetim sistemleri (ÖYS’ler) yeni özellikler ve servis seçenekleri kazanmıştır. Buna bağlı olarak artan ürün alternatifleri arasından seçim yapma süreci zorlaşmıştır. Belli kriterlere bağlı en uygun bulut tabanlı ÖYS’yi seçmek kurumlar için önemli bir karar verme sorunu olmuştur. Bu çalışmada, kurumların bir grup bulut tabanlı ÖYS arasından belli kriterlere uygun seçim yapabilmesini kolaylaştıracak Bulanık Analitik Hiyerarşi Süreci (BAHS) tabanlı bir model geliştirilmiştir. Bu modelde, bulut tabanlı ÖYS seçiminde etkili olabilecek içerik desteği, etkileşim ve iş birliği, ölçme ve değerlendirme, ders yapısı, arayüz, verimlilik araçları, platform esnekliği, ölçeklenebilirlik, güvenlik, destek ve lisanslama kriterleri literatüre ve uzman görüşlerine dayalı incelenmiş ve oluşturulan bir hiyerarşik yapı ile sunulmuştur. Çalışmada hem kriterler hem de durum çalışması kapsamında ele alınan altı alternatif, çevrim-içi eğitim alanında uzman yedi karar verici tarafından değerlendirilmiştir. Belirlenen kriterlere bağlı olarak, seçilen alternatifler arasında yapılan bulanık ikili karşılaştırmalar sonucu en uygun bulut tabanlı ÖYS, TalentLMS olarak belirlenmiştir.
Integration search strategies in tree seed algorithm for high dimensional function optimization
(SPRINGER HEIDELBERG, 2020) Gungor, Imral; Emiroglu, Bulent Gursel; Cinar, Ahmet Cevahir; Kiran, Mustafa Servet
The tree-seed algorithm, TSA for short, is a new population-based intelligent optimization algorithm developed for solving continuous optimization problems by inspiring the relationship between trees and their seeds. The locations of trees and seeds correspond to the possible solutions of the optimization problem on the search space. By using this model, the continuous optimization problems with lower dimensions are solved effectively, but its performance dramatically decreases on solving higher dimensional optimization problems. In order to address this issue in the basic TSA, an integration of different solution update rules are proposed in this study for solving high dimensional continuous optimization problems. Based on the search tendency parameter, which is a peculiar control parameter of TSA, five update rules and a withering process are utilized for obtaining seeds for the trees. The performance of the proposed method is investigated on basic 30-dimensional twelve numerical benchmark functions and CEC (congress on evolutionary computation) 2015 test suite. The performance of the proposed approach is also compared with the artificial bee colony algorithm, particle swarm optimization algorithm, genetic algorithm, pure random search algorithm and differential evolution variants. Experimental comparisons show that the proposed method is better than the basic method in terms of solution quality, robustness and convergence characteristics.
The Picard and Gauss-Weierstrass Singular Integrals in (p, q)-Calculus
(MALAYSIAN MATHEMATICAL SCIENCES SOC, 2020) Aral, A.; Deniz, E.; Erbay, H.
The vast development of the techniques in both the quantum calculus and the post-quantum calculus leads to a significant increase in activities in approximation theory due to applications in computational science and engineering. Herein, we introduce (p, q)-Picard and (p, q)-Gauss-Weierstrass integral operators in terms of the (p, q)-Gamma integral. We give a general formula for the monomials under both (p, q)-Picard and (p, q)-Gauss-Weierstrass operators as well as some special cases. We discuss the uniform convergence properties of them. We show that both operators have optimal global smoothness preservation property via usual modulus of continuity. Finally, we establish the rate of approximation using the weighted modulus of smoothness. Depending on the choices of parameters p and q in the integrals, we are able to obtain better error estimation than classical ones.
Detection of rheumatoid arthritis from hand radiographs using a convolutional neural network
(SPRINGER LONDON LTD, 2020) Ureten, Kemal; Erbay, Hasan; Maras, Hadi Hakan
Introduction Plain hand radiographs are the first-line and most commonly used imaging methods for diagnosis or differential diagnosis of rheumatoid arthritis (RA) and for monitoring disease activity. In this study, we used plain hand radiographs and tried to develop an automated diagnostic method using the convolutional neural networks to help physicians while diagnosing rheumatoid arthritis. Methods A convolutional neural network (CNN) is a deep learning method based on a multilayer neural network structure. The network was trained on a dataset containing 135 radiographs of the right hands, of which 61 were normal and 74 RA, and tested it on 45 radiographs, of which 20 were normal and 25 RA. Results The accuracy of the network was 73.33% and the error rate 0.0167. The sensitivity of the network was 0.6818; the specificity was 0.7826 and the precision 0.7500. Conclusion Using only pixel information on hand radiographs, a multi-layer CNN architecture with online data augmentation was designed. The performance metrics such as accuracy, error rate, sensitivity, specificity, and precision state shows that the network is promising in diagnosing rheumatoid arthritis.
Blind video quality assessment via spatiotemporal statistical analysis of adaptive cube size 3D-DCT coefficients
(INST ENGINEERING TECHNOLOGY-IET, 2020) Cemiloglu, Enes; Yilmaz, Gokce Nur
There is an urgent need for a robust video quality assessment (VQA) model that can efficiently evaluate the quality of a video content varying in terms of the distortion and content type in the absence of the reference video. Considering this need, a novel no reference (NR) model relying on the spatiotemporal statistics of the distorted video in a three-dimensional (3D)-discrete cosine transform (DCT) domain is proposed in this study. While developing the model, as the first contribution, the video contents are adaptively segmented into the cubes of different sizes and spatiotemporal contents in line with the human visual system (HVS) properties. Then, the 3D-DCT is applied to these cubes. Following that, as the second contribution, different efficient features (i.e. spectral behaviour, energy variation, distances between spatiotemporal frequency bands, and DC variation) associated with the contents of these cubes are extracted. After that, these features are associated with the subjective experimental results obtained from the EPFL-PoliMi video database using the linear regression analysis for building the model. The evaluation results present that the proposed model, unlike many top-performing NR-VQA models (e.g. V-BLIINDS, VIIDEO, and SSEQ), achieves high and stable performance across the videos with different contents and distortions.
Detection of breast cancer via deep convolution neural networks using MRI images
(SPRINGER, 2020) Yurttakal, Ahmet Hasim; Erbay, Hasan; Ikizceli, Turkan; Karacavus, Seyhan
Breast cancer is the type of cancer that develops from cells in the breast tissue. It is the leading cancer in women. Early detection of the breast cancer tumor is crucial in the treatment process. Mammography is a valuable tool for identifying breast cancer in the early phase before physical symptoms develop. To reduce false-negative diagnosis in mammography, a biopsy is recommended for lesions with greater than a 2% chance of having suspected malignant tumors and, among them, less than 30 percent are found to have malignancy. To decrease unnecessary biopsies, recently, Magnetic Resonance Imaging (MRI) has also been used to diagnose breast cancer. MRI is the highly recommended test for detecting and monitoring breast cancer tumors and interpreting lesioned regions since it has an excellent capability for soft tissue imaging. However, it requires an experienced radiologist and time-consuming process. On the other hand, convolutional neural networks (CNNs) have demonstrated better performance in image classification compared to feature-based methods and show promising performance in medical imaging. Herein, CNN was employed to characterize lesions as malignant or benign tumors using MRI images. Using only pixel information, a multi-layer CNN architecture with online data augmentation was designed. Later, the CNN architecture was trained and tested. The accuracy of the network is 98.33% and the error rate 0.0167. The sensitivity of the network is 1.0 whereas specificity is 0.9688. The precision is 0.9655.
Bibliometric analysis of publications on house dust mites during 1980-2018
(ELSEVIER ESPANA SLU, 2020) Demir, E.; Akmese, O. F.; Erbay, H.; Taylan-Ozkan, A.; Mumcuoglu, K. Y.
Background: The global prevalence of allergic diseases has increased dramatically in recent years and are now recognized as significant chronic diseases worldwide. One of the most important allergens that causes allergic diseases is house dust mites. Objective: This study aims to present a bibliometric overview of research published on dust mites between 1980 and 2018. Methods: Articles published from 1980 to 2018 were analyzed using bibliometric methods. The keywords ?Dust mite*,? and ? Dermatophagoides ? were used in the Web of Science (WoS). Simple linear regression analysis was used to estimate the number of future publications on this subject. Results: A total of 4742 publications were found, 2552 (53.8%) of them were articles. Most of the articles were on subjects related to immunology (1274; 49.9%) and allergy (1229; 48.1%). Clinical and Experimental Allergy (222; 8.7%) was the journal with the most publications. The USA was the country that most contributed to the literature with 461 (18.1%) articles. The countries producing the most publications on this subject were developed countries. The most active author was W.R. Thomas (66; 2.5%). The most productive institution was the University of Western Australia (91; 3.6%). The most cited article was published in the New England Journal of Medicine .Conclusion: According to the findings, developed countries were the most productive in publishing on house dust mites. By planning multinational research rather than regional studies, it may be suggested that researchers in underdeveloped or developing countries could also-conduct more research on this subject.(C) 2019 SEICAP. Published by Elsevier Espana, S.L.U. All rights reserved.
Android malware detection based on image-based features and machine learning techniques
(SPRINGER INTERNATIONAL PUBLISHING AG, 2020) Unver, Halil Murat; Bakour, Khaled
In this paper, a malware classification model has been proposed for detecting malware samples in the Android environment. The proposed model is based on converting some files from the source of the Android applications into grayscale images. Some image-based local features and global features, including four different types of local features and three different types of global features, have been extracted from the constructed grayscale image datasets and used for training the proposed model. To the best of our knowledge, this type of features is used for the first time in the Android malware detection domain. Moreover, the bag of visual words algorithm has been used to construct one feature vector from the descriptors of the local feature extracted from each image. The extracted local and global features have been used for training multiple machine learning classifiers including Random forest, k-nearest neighbors, Decision Tree, Bagging, AdaBoost and Gradient Boost. The proposed method obtained a very high classification accuracy reached 98.75% with a typical computational time does not exceed 0.018 s for each sample. The results of the proposed model outperformed the results of all compared state-of-art models in term of both classification accuracy and computational time.
VisDroid: Android malware classification based on local and global image features, bag of visual words and machine learning techniques
(SPRINGER LONDON LTD, 2020) Bakour, Khaled; Unver, Halil Murat
In this paper, VisDroid, a novel generic image-based classification method has been suggested and developed for classifying the Android malware samples into its families. To this end, five grayscale image datasets each of which contains 4850 samples have been constructed based on different files from the contents of the Android malware samples sources. Two types of image-based features have been extracted and used to train six machine learning classifiers including Random Forest, K-nearest neighbour, Decision trees, Bagging, AdaBoost and Gradient Boost classifiers. The first type of the extracted features is local features including Scale-Invariant Feature Transform, Speeded Up Robust Features, Oriented FAST and Rotated BRIEF (ORB) and KAZE features. The second type of the extracted features is global features including Colour Histogram, Hu Moments and Haralick Texture. Furthermore, a hybridized ensemble voting classifier has been proposed to test the efficiency of using a number of machine learning classifiers trained using local and global features as voters to make a decision in an ensemble voting classifier. Moreover, two well-known deep learning model, i.e. Residual Neural Network and Inception-v3 have been tested using some of the constructed image datasets. Furthermore, when the results of the proposed model have been compared with the results of some state-of-art works it has been revealed that the proposed model outperforms the compared previous models in term of classification accuracy, computational time, generality and classification mode.
Context-dependent model for spam detection on social networks
(SPRINGER INTERNATIONAL PUBLISHING AG, 2020) Ghanem, Razan; Erbay, Hasan
Social media platforms are getting an important communication medium in our daily life, and their increasing popularity makes them an ideal platform for spammers to spread spam messages, known as spam problems. Moreover, messages on social media are vague and messy, so a good representation of the text may be the first step to address spam problem. While traditional weighting methods suffer from both high dimensionality and high sparsity problems, traditional word embedding methods suffer from context independence and out of vocabulary problems. To overcome these problems, in this study, we propose a novel architecture based on a context-dependent representation of text using the BERT model. The model was tested using the Twitter dataset, and experimental results show that the proposed method outperforms traditional weighting methods, traditional word embedding based methods as well as the existing state of the art methods used to detect spam on the twitter platform.

Güncel Gönderiler