Yazar "Adem, Kemal" seçeneğine göre listele
Listeleniyor 1 - 14 / 14
Sayfa Başına Sonuç
Sıralama seçenekleri
Öğe A hybrid model based on the convolutional neural network model and artificial bee colony or particle swarm optimization-based iterative thresholding for the detection of bruised apples(Türkiye Bilimsel ve Teknolojik Araştırma Kurumu (TÜBİTAK), 2020) Hekim, Mahmut; Cömert, Onur; Adem, KemalIn this study, apple images taken with near-infrared (NIR) cameras were classified as bruised and healthy objects using iterative thresholding approaches based on artificial bee colony (ABC) and particle swarm optimization (PSO) algorithms supported by a convolutional neural network (CNN) deep learning model. The proposed model includes the following stages: image acquisition, image preprocessing, the segmentation of anatomical regions (stem-calyx regions) to be discarded, the detection of bruised areas on the apple images, and their classification. For this aim, by using the image acquisition platform with a NIR camera, a total of 1200 images at 6 different angles were taken from 200 apples, of which 100 were bruised and 100 healthy. In order to increase the success of detection and classification, adaptive histogram equalization (AHE), edge detection, and morphological operations were applied to the images in the preprocessing stage, respectively. First, in order to segment and discard the stem-calyx anatomical regions of the images, the CNN model was trained by using the preprocessed images. Second, the threshold value was determined by means of the ABC/PSO-based iterative thresholding approach on the images whose stem-calyx regions were discarded, and then the bruised areas on the images with no stem-calyx anatomical regions were detected by using the determined threshold value. Finally, the apple images were classified as bruised and healthy objects by using this threshold value. In order to illustrate the classification success of our approaches, the same classification experiments were reimplemented by directly using the CNN model alone on the preprocessed images with no ABC and PSO approaches. Experimental results showed that the hybrid model proposed in this paper was more successful than the CNN model in which ABC-and PSO-based iterative threshold approaches were not used.Öğe Automatic detection and classification of leaf spot disease in sugar beet using deep learning algorithms(Elsevier B.V., 2019) Özgüven, Mehmet Metin; Adem, KemalDepending on the severity of the leaf spot disease in the field, it can cause a loss in sugar yield by 10% to 50%. Therefore, disease symptoms should be detected on-time and relevant measures should be taken instantly to prevent further spread or progress of the disease. In this study, an Updated Faster R-CNN architecture developed by changing the parameters of a CNN model and a Faster R-CNN architecture for automatic detection of leaf spot disease (Cercospora beticola Sacc.) in sugar beet were proposed. The method, proposed for the detection of disease severity by imaging-based expert systems, was trained and tested with 155 images and according to the test results, the overall correct classification rate was found to be 95.48%. In addition, the proposed approach showed that changes in CNN parameters according to the image and regions to be detected could increase the success of Faster R-CNN architecture. The proposed approach yielded better outcomes for relevant parameters than the modern methods specified in previous literature. Therefore, it is believed that the method will reduce the time spent in diagnosis of sugar beet leaf spot disease in the large production areas as well as reducing the human error and time to identify the severity and course of the disease.Öğe COVID-19 diagnosis prediction in emergency care patients using convolutional neural network(Afyon Kocatepe Üniversitesi, 2021) Adem, Kemal; Kılıçarslan, SerhatThe sudden increase in cases of Coronavirus disease (COVID-19) puts a high pressure on health care providers in many countries across the world. In the present case, an early and correct diagnosis of the disease, and starting the treatment is of vital importance. Most of the developing countries have insufficient RT-PCR tests, the most verified diagnostic test for COVID-19. This increases the number of infected patients and delays preventive measures. In this study, the risk of a positive COVID-19 diagnosis is estimated by applying Convolutional Neural Network (CNN) method, which is a deep learning model, to the dataset obtained from routine blood tests of all patients who admitted to the emergency service. The dataset used in the experiments consists of the data from patients admitted to the Israelita Albert Einstein Hospital in São Paulo, Brazil, between March 28th and April 3rd, 2020. In addition to the J48, ANN, Random Forest, and Random Committee data mining algorithms, the CNN deep learning algorithm were applied to the dataset. The 5 and 7 fold cross validation model was applied to the data set and the average of the two models was used as an evaluation criterion in order to ensure objectivity.The best prediction performance was obtained by the CNN method by 92.52% accuracy. Experimental results revealed that the proposed approach is in line with the results of the tests with general validity.Öğe Covid19 yayılımını azaltmak için yüz maskesinin evrişimsel sinir ağı modelleri ile tespiti(Iğdır Üniversitesi, 2023) Daşgın, Aslıhan; Adem, Kemal; Kılıçarslan, SerhatSon yıllarda hayatımızın gerçeği olan ve tüm dünya için pandemi haline gelen Covid-19’un gerekli tedbirlere sıkı şekilde uyulmadığı takdirde bulaşma oranı artmakta hatta varyantları bile ortaya çıkmaya başlamaktadır. DSÖ tarafından yayınlanan ve alınması gerekli olan tedbirler alındıkça hastalıkla mücadele daha kolay hale gelebilmektedir. Tedbirlere uymanın zorluğu olsa da uymaya özen gösterildiği taktirde, hastalık ya daha hafif atlatılmakta ya da hastalığa kolayca yakalanılmamaktadır. Bu tedbirlerin en önemlilerinden birisi de kalabalık alanlarda maske kullanımına özen gösterilmesidir. Maske kullanımının önemi araştırmalarla desteklenmesinden sonra, bazı alışveriş merkezi, sağlık kuruluşları, okullar gibi kalabalık mekanlarda maske kullanımına yönelik denetimler başlamıştır. Ancak bu denetimleri bir insanın gerçekleştirmesi zor olduğundan günümüzde sıklıkla kullanılan derin öğrenme yöntemleriyle maske tespiti çalışmaları yapılmaya başlanmıştır. Bu tez çalışmasında, transfer öğrenme tabanlı modeller kullanılarak maske tespiti gerçekleştirilmesi amaçlanmaktadır. Kaggle web sitesinde bulunan veri seti ile toplamda 906 görüntü ile DenseNet121, EfficientNetV2M, NasNetMobile, InceptionV3, VGG19 ve InceptionResNetV2 derin öğrenme modelleri kullanılmıştır. Deneysel değerlendirmeler sonucunda, en iyi başarı oranı olarak NasNetMobile modeliyle, %99.35 doğruluk, %99 kesinlik, %99 geri çağırma ve %99 f1 skorları elde edildiği görülmüştür.Öğe Defect detection of seals in multilayer aseptic packages using deep learning(Türkiye Klinikleri, 2019) Adem, Kemal; Közkurt, CemilSealing in aseptic packages, one of the healthiest and cheapest technologies to protect food from parasites in the liquid food industry, requires a detailed and careful control process. Since the controls are made manually and visually by expert machine operators, the human factor can lead to the failure to detect defects, resulting in high cost and food safety risks. Therefore, this study aims to perform a leak test in aseptic package seals by a system that makes decisions using independent deep learning methods. The proposed Faster R-CNN and the Updated Faster R-CNN deep learning models were subjected to training and testing with a total of 400 images taken from a real production environment, resulting in a correct classification rate of 99.25%. As a result, it can be said that the study is the second study that performs a computer-aided quality control process with promising results, having distinctive features such as being the first study that conducts analysis using the deep learning methodÖğe Derin öğrenme ve makine öğrenmesi yöntemleri kullanılarak gelişmekte olan ülkelerin finansal enstrümanlarının etkileşimi ile Bist 100 tahmini(Niğde Ömer Halisdemir Üniversites, 2023) Akbulut, Serap; Adem, KemalDöviz piyasaları, emtia piyasaları ve gelişmekte olan ülkelerin borsa endekslerinin Bist100 üzerindeki etkisi oldukça önemlidir. Ülke ekonomileri hem kendi hem de diğer ülkelerin ekonomilerine güçlü bir şekilde bağlıdır ve bu sebeple piyasalar etkilenmektedirler. Ekonomik açıdan piyasayı takip etmek kararları doğru verebilmek için tahmin yöntemleri kullanılmaktadır. Veri kümesi Ocak 2017 – Ekim 2021 tarihleri arasındaki kapanış verilerinden oluşmaktadır. Deneysel çalışmalarda objektifliğin sağlanması amacıyla k=5 ve 10 katlı çapraz geçerlilik modeli uygulanmıştır. Modellerin karşılaştırılmasında; Ortalama Mutlak Hata (MAE), Bağıl Mutlak Hata (RAE), Ortalama Karesel Hata Karekökü (RMSE) ve uzun kısa süreli bellek (LSTM) kullanılmıştır. Deneysel çalışmalar sonucunda, LSTM modelinin makine öğrenmesi modellerine göre daha iyi sonuç verdiği görülmüştür. LSTM modeli için test sonuçları incelendiğinde MAE değeri 10.27, RMSE değeri 14.15 ve RAE değeri ise 6.06’dir.Öğe Detection of Monkeypox disease from skin lesion images using deep learning methods(Niğde Ömer Halisdemir Üniversitesi, 2024) Engin, Muhammet Talha; Adem, KemalMonkeypox is a disease that, while less deadly and contagious than COVID-19, could pose a global pandemic threat. In the field of medical imaging, deep learning techniques offer promising results in the diagnosis of diseases. This study develops deep learning models using skin lesion images for early diagnosis of monkeypox. The research is divided into two key sections. In the first section, a deep learning model is developed using the Monkeypox Skin Image Dataset (MSID). The second section focuses on a model trained on a combined dataset, which merges the Monkeypox Skin Image, Monkeypox Master, and Monkeypox Original Image Datasets, referred to as HYBRID. The MSID dataset comprises 806 Monkeypox and 690 Non-Monkeypox images for training, along with 309 Monkeypox and 292 Non-Monkeypox images for testing, resulting in a total of 2,097 images of skin lesions with and without monkeypox. The HYBRID dataset includes 1,088 Monkeypox and 1,896 Non-Monkeypox images for training, as well as 468 Monkeypox and 812 Non-Monkeypox images for testing, resulting in a total of 4,264 skin lesion images. Five distinct deep learning models—DenseNet201, InceptionResNetV2, InceptionV3, NASNetLarge, and Xception—were applied to both datasets, and the outcomes were compared. The DenseNet201 model, when trained on augmented data, demonstrated remarkable performance in detecting monkeypox, achieving accuracy rates of 99.33% on the MSID dataset and 98.52% on the HYBRID dataset.Öğe Diagnosis and classification of cancer using hybrid model based on ReliefF and convolutional neural network(Churchill Livingstone, 2020) Kılıçarslan, Serhat; Adem, Kemal; Çelik, MeteMachine learning and deep learning methods aims to discover patterns out of datasets such as, microarray data and medical data. In recent years, the importance of producing microarray data from tissue and cell samples and analyzing these microarray data has increased. Machine learning and deep learning methods have been started to use in the diagnosis and classification of microarray data of cancer diseases. However, it is challenging to analyze microarray data due to the small number of sample size and high number of features of microarray data and in some cases some features may not be relevant with the classification. Because of this reason, studies in the literature focused on developing feature selection/dimension reduction techniques and classification algorithms to improve classification accuracy of the microarray data. This study proposes hybrid methods by using Relief and stacked autoencoder approaches for dimension reduction and support vector machines (SVM) and convolutional neural networks (CNN) for classification. In the study, three microarray datasets of Overian, Leukemia and Central Nervous System (CNS) were used. Ovarian dataset contains 253 samples, 15,154 genes and 2 classes, Leukemia dataset contains 72 samples, 7129 genes, and 2 classes and CNS dataset contains 60 samples, 7129 genes and 2 classes. Among the methods applied to the three microarray data, the best classification accuracy without dimension reduction was observed with SVM as 96.14% for ovarian dataset, 94.83% for leukemia dataset and 65% for CNS dataset. The proposed hybrid method ReliefF + CNN method outperformed other approaches. It gave 98.6%, 99.86% and 83.95% classification accuracy for the datasets of ovarian, leukemia, and CNS datasets, respectively. Results shows that dimension reduction methods improved the classification accuracy of the methods of SVM and CNN.Öğe Diagnosis of breast cancer with Stacked autoencoder and Subspace kNN(Elsevier B.V., 2020) Adem, KemalBreast cancer is one of the most common and deadliest cancer types in women worldwide. Research on this disease has become very important because early diagnosis stages, clinical applications and the speed of response to treatment are facilitated in diseases such as cancer. In this study, an approach is proposed in which a Subspace kNN algorithm is used together with Stacked autoencoder (SAE) for diagnosis of disease on the breast cancer microarray dataset for the first time. Such hybrid approaches can provide better results when classifying data sets with high-dimensional and uncertainty. The data set used in the study was taken from Kent Ridge-2 database. It consists of 97 samples (51 benign, 46 malicious) and 24482 attributes. The performance of the proposed method was evaluated and the results were compared with other well-known methods of dimension reduction and machine learning. As a result of the comparison, the data set was reduced to 100 attributes by using SAE and Subspace kNN and 91.24% accuracy was achieved. The result obtained provides important classification accuracy, especially in high-dimensional data sets. The importance of this study is that the models that were created by using various classifiers to increase the success rate of the stacked autoencoder-softmax classifier model in the breast cancer microarray data set were applied for the first time. In this regard, it is considered that automation-based studies will provide diagnostic decision support system a solution using the proposed method in future works.Öğe Impact of activation functions and number of layers on detection of exudates using circular Hough transform and convolutional neural networks(Elsevier, 2022) Adem, KemalConvolutional neural networks (CNN), which are used for object detection, are widely used in image recognition and segmentation fields due to their good performance. In this study, a method is proposed for the detection of exudate from DR lesions based on image processing and CNN model with different activation functions and layer numbers. Activation functions are one of the main reasons why the CNN model is complicated in hierarchical structure. CNN models have real artificial intelligence capabilities, especially thanks to non-linear activation functions. Although ReLU is the most commonly used activation function in CNN models, it has some shortcomings in practice. The most important shortcoming is that if the input value is negative, the learning process will slow down due to the inability to take the derivative. In order to solve this problem, the effect of the activation function in the CNN model on a real world problem and its image was investigated in this study. By applying circular Hough transform, one of the image processing methods, the OD region, which has a similar structure to the exuding regions, was determined and removed from the image, and then the detection of the exuding regions was carried out using the CNN model. Exudate detection was performed using ReLU, ELU, Leaky ReLU, Softplus and Swish activation functions in the CNN model and the results were compared. Experimental results in DiaretDB0 and DiaretDB1 databases show that the Swish activation function has a better performance than the others.Öğe Optimizing hyperparameters for enhanced performance in convolutional neural networks: A study using NASNetMobile and DenseNet201 Models(Bandırma Onyedi Eylül Üniversitesi, 2024) Aksoy, İbrahim; Adem, KemalConvolutional neural networks, inspired by the workings of biological neural networks, have proven highly successful in tasks like image data recognition, classification, and feature extraction. Yet, designing and implementing these networks pose certain challenges. One such challenge involves optimizing hyperparameters tailored to the specific model, dataset, and hardware. This study delved into how various hyperparameters impact the classification performance of convolutional neural network models. The investigation focused on parameters like the number of epochs, neurons, batch size, activation functions, optimization algorithms, and learning rate. Using the Keras library, experiments were conducted using NASNetMobile and DenseNet201 models—highlighted for their superior performance on the dataset. After running 65 different training sessions, accuracy rates saw a notable increase of 6.5% for NASNetMobile and 11.55% for DenseNet201 compared to their initial values.Öğe Performance analysis of optimization algorithms on stacked autoencoder(Institute of Electrical and Electronics Engineers Inc., 2019) Adem, Kemal; Kılıçarslan, SerhatStacked autoencoder (SAE) model, which is one of the deep learning methods, has been widely used in one dimensional data sets in recent years. In this study, a comparative performance analysis was performed using the five most commonly used optimization techniques and two well-known activation functions in SAE architecture. Stochastic Gradient Descent (SGD), Root Mean Square Propagation (RmsProp), Adaptive Moment Estimation (Adam), Adaptive Delta (Adadelta) and Nesterov-accelerated Adaptive Moment Estimation (Nadam) and Softmax and Sigmoid were used as optimization techniques. In this study, two different data sets in public UCI database were used. In order to verify the performance of the SAE model, experimental studies were performed by using the obtained data sets together with optimization and activation techniques separately. As a result of the experimental studies, the success rate of 88.89%, 85.19% in Cryotherapy and Immunotherapy data set was achieved by using Softmax activation function with SGD optimization method on three-layer SAE. After a successful training phase, adaptive optimization techniques Adam, Adadelta, Nadam and RmsProp methods were observed to have a weaker learning process than the stochastic method SGD.Öğe RSESLIBKNN makine öğrenmesi yöntemi kullanılarak parkinson hastalığının tanısı(Niğde Ömer Halisdemir Üniversitesi, 2020) Bütüner, İlknur; Kaplan, Burak; Adem, KemalParkinson hastalığı, insanların yaşam kalitesini etkileyen nörolojik bir hastalıktır. Parkinson hastalığı merkezi sinir sistemini olumsuz etkileyen bir hastalıktır. Hastaların ölümüne yol açabilmektedir. Bu nedenle, Parkinson hastalığının erken tespiti son derece önemlidir. Parkinson hastalığına ait belirtiler, potansiyel olarak gelişmiş makine öğrenme tekniklerine dayanan bilgisayar destekli tanı sistemleri ile tespit edilebilir. Bu çalışmada Parkinson hastalığı tanısı için kNN, RseslibKnn ve A1DE makine öğrenmesi yöntemleri kullanılmıştır. Çalışmanın amacı Parkinson hastalığı veri kümesi üzerinde makine öğrenmesi yöntemlerinin başarı oranlarının karşılaştırılarak en uygun karar destek sisteminin sunulmasıdır. Veri kümesi olarak ‘UC Irvine Machine Learning Repository’ veri tabanından elde edilen, 252 örnekten ve 753 öznitelikten oluşan veri kümesi kullanılmıştır. Literatür üzerinde farklı çalışmalar da incelenip karşılaştırılmıştır. Deneysel çalışmalar farklı çapraz geçerlilikler üzerinden yapılmış olup bunların ortalaması başarı sonucu olarak verilmiştir. Çalışma sonucunda, parkinson hastalığı veri kümesi kNN, RseslibKnn ve A1DE makine öğrenmesi yöntemleri ile sınıflandırılmış ve daha sonra eğitim ve test sonuçları doğruluk, duyarlılık ve özgüllük değerleri temel alınarak değerlendirilmiştir. Farklı çapraz geçerlilik değerleri ile ele alınan tüm yöntemler incelediğinde en yüksek başarı sonucu %97,61 doğruluk oranı ortalaması ile RseslibKnn yöntemi vermiştir. Değerlendirme sonucunda RseslibKnn makine öğrenmesi yönteminin Parkinson hastalığının tespiti ile ilgili karar destek sistemleri üzerine önerilerde bulunulmuştur.Öğe The assessment of different bleaching agents’ efficiency on discoloured teeth using image-processing methods(Elsevier B.V., 2020) Özkoçak, İsmail; Hekim, Mahmut; Göktürk, Hakan; Adem, Kemal; Cömert, OnurBackground: Although triple antibiotic paste (TAP) has been successfully used as an intracanal medicament for regenerative endodontic treatments, TAP has also been shown to cause discolouration. The aim of this study is to investigate the efficacy of different bleaching agents to bleach teeth discoloured from TAP. Methods: Two hundred extracted human maxillary incisors were evaluated with VITA Easyshade, and 120 teeth were prepared and discoloured by using TAP for three weeks. After colouration, 70 teeth were randomly divided into five groups: Group 1: Negative control, Group 2: Sodium perborate, Group 3: Opalescence Endo, Group 4: Endoperox, and Group 5: Biolase. The colour changes in the third and seventh days’ standard images were obtained using stereomicroscopy, RGB and Lab color space transformations were applied to the images. The CIE Lab color system was used, and total color changes (?E) were calculated and compared among groups and over time, using analysis of variance testing. Results: At the third day, there was no difference between bleaching materials. At the seventh day, the Biolase group was superior to sodium perborate and there was no difference between other groups. A statistically significant difference was noted between the third and seventh-day measurements for all bleaching protocols. Bleaching effectiveness of all agents increased over time. Conclusions: Teeth discoloured by using TAP may be bleached by means of the investigated protocols, and colour alteration can be increased over time. The CIE Lab colour system can be used as an alternative, in vitro test for evaluating the bleaching efficiency of bleaching agents.
