Search

Dijital dönüşüm ile tedarik zinciri yönetiminde büyük veri analitiğinin önemi gün geçtikçe artmaktadır. Özellikle müşteri taleplerinin hızlı ve doğru tahmin edilmesinde büyük verinin kullanımı firmalara rekabet avantajı sağlamaktadır. Bu doğrultuda, yapay zekâ tekniklerinden biri olan derin öğrenme modelleri büyük verideki karmaşık örüntülerin keşfedilmesinde öne çıkmaktadır. Son yıllarda literatürde çok sayıda derin öğrenme yöntemi önerilmiştir. Bu çalışmada, satış tahmini problemi için derin öğrenme yöntemlerinin performansları karşılaştırılmıştır. Bu kapsamda derin sinir ağı (DNN), derin otokodlayıcı (Deep AE), evrişimli sinir ağı (CNN), tekrarlayan sinir ağı (RNN), uzun kısa-süreli bellek (LSTM) ağı, çift yönlü LSTM (Bi-LSTM) ağı, kapılı tekrarlayan birim (GRU), CNN-LSTM ve evrişimli LSTM (ConvLSTM) yöntemleri uygulanmıştır. Çeşitli sektörlere ait satış verileri kullanılarak deneysel çalışmalar gerçekleştirilmiştir. Hiperparametre optimizasyonu ardından ele alınan yöntemlerin performansları tahmin doğruluğu ve eğitim süreleri açısından karşılaştırılarak sonuçların istatistiksel anlamlılığı değerlendirilmiştir. Sonuç olarak, LSTM ve GRU modellerinin tahmin doğruluğunda başarılı sonuçlar verdiği, CNN modelinin ise eğitim süresini kısalttığı görülmüştür.

Ensuring security in speaker recognition systems is crucial. In the past years, it has been demonstrated that spoofing attacks can fool these systems. In order to deal with this issue, spoof speech detection systems have been developed. While these systems have served with a good performance, their effectiveness tends to degrade under noise. Traditional speech enhancement methods are not efficient for improving performance, they even make it worse. In this research paper, performance of the noise mask obtained via a convolutional neural network structure for reducing the noise effects was investigated. The mask is used to suppress noisy regions of spectrograms in order to extract robust i-vectors. The proposed system is tested on the ASVspoof 2015 database with three different noise types and accomplished superior performance compared to the traditional systems. However, there is a loss of performance in noise types that are not encountered during training phase.

Copyright of Bingol University Journal of Economics & Administrative Science is the property of Bingol University Journal of Economics & Administrative Science and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Karaelmas Science & Engineering Journal / Karaelmas Fen ve Mühendislik Dergisi is the property of Karaelmas Science & Engineering Journal and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Plant disease classification is the use of machine learning techniques for determining the type of disease from the input leaf images of the plants based on certain features. It is an important research area since early identification and treatment of plant disease is critical for saving crops, preventing agricultural disasters, and improving productivity in agriculture. This study proposes a new convolutional neural network model that accurately classifies the diseases on the plant leaves for the agriculture sectors. It especially works on the classification of plant diseases for grape leaves from images by designing a deeplearning architecture. A web application was also implemented to help the agricultural workers. The experiments carried out on real-world images showed that a significant improvement (8.7%) on average was achieved by the proposed model (98.53%) against the state-of-the-art models (89.84%) in terms of accuracy.

Recognition of handwritten digits has recently gained importance and attracted the attention of many scientists, as it is used in many machine learning, deep learning and computer vision applications. Hyperparameter optimization involves determining a set of values aimed at increasing accuracy in both classification and prediction. It is also aimed to optimize the performance in feature selection by regulating the parameters selected by the algorithms more accurately. In this study, a convolutional neural network was used to recognize handwritten digits using the MNIST dataset. There are many open source hyperparameter libraries that deep learning developers can use to determine hyperparameters. In the developed model, hyperparameter optimization techniques were applied using Optuna, HyperOpt and Scikitoptimize libraries and the results were evaluated. Optimization times for hyperparameter libraries and the change in the success rate in recognizing handwritten digits were analyzed. The model trained with randomly given parameters achieved 78.45%, 97.13%, 75.62%, 76.95%, 97.46% and 97.27% accuracy, while the model trained with optimized hyperparameters achieved 99.26% accuracy. [ABSTRACT FROM AUTHOR]

Copyright of SDU Journal of Engineering Sciences & Design / Mühendislik Bilimleri ve Tasarım Dergisi is the property of Journal of Engineering Sciences & Design and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Classification studies on image data sets with convolutional neural networks algorithm, which is one of the deep learning approaches, are widely performed successfully in many fields such as medicine and agriculture. However, in cases where the sample numbers of the classes in the image datasets are imbalanced, the classification performance of this algorithm is negatively affected. In general, unlike the majority class, the minority class(es) are not well learned by the convolutional neural network algorithm. In such cases, applying oversampling methods provides successful results. With oversampling methods, the number of minority class(s) samples is increased, making it close to or equal to the sample number of the majority class. In this study, translation, rotation, random erasing, noise injection, mixing of images, kernel filters, generative adversarial networks, flipping, feature space transformation, cropping and color space transformation oversampling methods frequently used in the literature were applied on the Ham10000 dataset. According to the results obtained as a result of the study, oversampling methods have been compared in terms of classification performance. From three different convolutional neural network models, according to the classification results obtained with ResNet50, DenseNet201, MobileNet, in terms of accuracy, the noise injection method in the ResNet50 model was 0.967, the color space transformation method in the DenseNet201 model was 0.965, and in the MobileNet model, the mixing of images method had a classification performance value of 0.974 which was better than the other oversampling methods. [ABSTRACT FROM AUTHOR]

Hızla gelişen bilgisayar ve grafik ara yüzüne sahip cihaz teknolojileri, yüz tanıma çalışmalarında yeni ufuklar açmışlardır. Özellikle derin öğrenme ağ mimari yapılarından biri olan evrişimsel sinir ağları (Convolutional Neural Network-CNN), yüz tanıma çalışmalarında büyük başarılar sağlamaktadır. Bu başarılar da veri setlerinin büyüklüğü önemli rol oynamaktadır. Özellikle kullanılan veri setlerindeki yetersizlik başarı oranlarını etkileyebilmektedir. Bunun önüne geçmek için ise veri tipine göre değişik veri artırma teknikleri uygulanmaktadır. Yapılan bu çalışmada yüz tanımlama problemi için derin öğrenmeye dayalı adaptif bir yüz tanıma modeli (AYTM) geliştirildi. Geliştirilen bu model kontrast sınırlı uyarlanabilir histogram eşitleme (Contrast Limited Adaptive Histogram Equalization-CLAHE), CNN ve çok katmanlı algılayıcı (Multi Layer Perceptron-MLP)’ndan oluşmaktadır. İki farklı veri seti grubu kullanılarak geliştirilen modelin performans değerlendirilmesi yapılmıştır. Özellikle veri artırma işleminin model başarısını ciddi oranda artırdığı gözlendi ve veri artırma işleminin derin öğrenme uygulamalarında gerekliliği vurgulanmıştır.

Copyright of International Journal of InformaticsTechnologies is the property of Institute of Informatics, Gazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Duzce University Journal of Science & Technology is the property of Duzce University Journal of Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The U-Net architecture, which performs segmentation on images, has also achieved very successful results in the medical field. However, there is also a need to improve the U-Net architecture for better results. In this article, some improvement proposals are presented for the U-Net model's encoder part, and the segmentation success of the implemented architecture for segmentation of dermoscopic image lesions is evaluated. The PH2 dataset and the "International Skin Imaging Collaboration" datasets (ISIC-2016 and ISIC-2017) were used for the research. The traditional data augmentation method was applied to the selected PH2 dataset samples. The results of the proposed model (EnecaU-Net) and the U-Net model obtained with the PH2 dataset were compared. Furthermore, in this article, the mix data augmentation method, which has an influence on the model's segmentation success, is examined for lesion segmentation on dermoscopic images. This investigation was made with the ISIC-2016 dataset, and its experimental results were compared with the same amount of the ISIC-2017 dataset that didn't apply data augmentation operations. Although, during the evaluation phase, Dice and Jaccard (IoU) metrics were used primarily to measure the success of the model, specificity, sensitivity, and accuracy criteria were also used. According to our results, the segmentation success of the EnecaU-Net model applied for lesion segmentation on dermoscopic images is high, and the applied mix data augmentation method improves the segmentation success of the EnecaU-Net model. The average test results achieved by the proposed model are 88.05% and 80.30% for ISIC2016 and 83.09% and 74.54% for ISIC-2017 in terms of the Dice and Jaccard values, respectively. [ABSTRACT FROM AUTHOR]

Developments in technology have significantly affected the marketing activities of businesses. With the integration of technology into marketing activities, sales increased and the attention of target markets began to be drawn more. Thanks to the opportunities brought by technology, businesses have had the chance to understand the personal needs and expectations of consumers more easily. Thus, personalization in marketing has begun to take place at the center of marketing. However, developments in recent years have brought the issue of hyper-personalization, which is one step beyond personalization of businesses, to the agenda. The effects of artificial intelligence, machine learning, internet of things have an important impact great in this. In this paper, gait-based gender recognition problem, which is an important example for hyper-personalized marketing activities, was attempted to solve with Convolutional Neural Networks (CNNs). Various networks underwent evaluation for this task, with one chosen as a foundation. Subsequent modifications were applied to this base network through experimentation with architectural choices and hyperparameters. Despite exhibiting promising performance akin to prior research, the experimental findings shed light on the impact of network structure and hyperparameters on performance. The experiments employed gait silhouette, a feature descriptor, as input. The resultant overall accuracy reached 97.45% with the proposed CNN architecture. This outcome offers valuable insights into utilizing gait feature descriptions for classification within the problem domain. [ABSTRACT FROM AUTHOR]

This paper proposes a novel approach for intrusion detection using a metaheuristic-based feature selection method combined with convolutional neural networks (CNNs). The feature selection method employs a decision tree and a metaheuristic algorithm to select the most important features from different datasets. The selected features are then feed into CNNs, including ResNet50, VGG16, and EfficientNet, to improve the accuracy of intrusion detection. Experimental results on several benchmark datasets show that the proposed method can be promising in terms of different criteria. The final results prove that EfficientNet and ResNet50 perform much better than VGG16. When EfficientNet and ResNet50 algorithms are applied to NSL-KDD, DEFCON and CDX datasets, the best accuracy rates are 96.2% and 81.3% correspondingly. In addition, while EfficientNet has the highest rate of 98.6% according to the specificity criterion, ResNet50 stands out with a recall rate of 95.1% and a rate of 95.2% for F1score. [ABSTRACT FROM AUTHOR]

Copyright of Duzce University Journal of Science & Technology is the property of Duzce University Journal of Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Konya Journal of Engineering Sciences / Konya Mühendislik Bilimleri Dergisi is the property of Selcuk University Journal of Engineering, Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In this study, InceptionV3 based Enriched Feature Integration Network (Inc-EFIN) architecture was developed for automatic detection of surface defects. In the proposed architecture, features of all levels of the InceptionV3 architecture are extracted and the features with the same height and width are combined. As a result of merging, 5 feature maps were obtained. Channel-Based Squeeze and Excitation block has been applied to reveal important details in these feature maps. In Feature Pyramid Network module, information from low-level feature maps containing spatial details were transferred to high-level feature maps containing semantic details. Then, for the final feature map, features were combined using the Feature Integration and Signification (FIS) module. The feature map combined in the FIS module was passed through the Spatial and Channel-based Squeeze and Excitation block. Defect detection results were obtained by using convolution and sigmoid layers in the last layer of the Inc-EFIN architecture. MT, MVTec-Texture, and DAGM datasets were used to calculate the pixel-level defect detection success of the Inc-EFIN architecture. In experimental studies, Inc-EFIN architecture achieved higher performance than the latest technologies in the literature with 77.44% mIoU, 81.2% mIoU and 79.46% mIoU performance results in MT, MVTec-Texture and DAGM datasets, respectively. [ABSTRACT FROM AUTHOR]

In this study, crack identification and detection process in sheet metals, which is one of the most important issues especially for sheet metal forming companies, was investigated. Cracks in sheet material, which are frequently seen in mass production, cause vehicle scraps, so the degree of impact and cost are high. Crack tests were performed on 490 different sheets on sheet material in the 0.6-3 mm thickness range, and a crack image dataset was created for sheet material. In addition, 121 non-cracked "accepted-approved" parts were included in the dataset. Due to its superior feature extraction capability, convolutional neural network (CNN) has been widely researched, applied and outperformed other traditional machine learning methods in the field of intelligent fault diagnosis. After that, the cracked and non-cracked part data in the sheet material was divided into training and test data, and the accuracy values. Here, accuracy and verification accuracy rates as 98.5% and 90% were achieved respectively. Accuracy values with VGGNet architecture are 98.75% and 90%. This gives information about the reliability of both the picture data of the faults and the model presented. The results of the research are important as they will form the basis for the detection of crack defects in sheet material, which is very important in mass production applications. [ABSTRACT FROM AUTHOR]

There has been a significant increase in the use of more than one drug, especially in recent years. Concomitant use of medications by a patient can lead to drug-drug interactions, which are defined as drug-to-drug reactions. In order to prevent adverse situations, it is necessary to predict the interactions that may occur between drugs. The prediction of drug-drug interactions is usually carried out with experiments and requires an intense workload. In order for clinicians to make more accurate decisions and create appropriate treatment programs, approaches that perform automatic prediction of drug-drug interaction have been frequently used in the literature. Although many studies have been developed in the literature for prediction of drug-drug interaction, there are still significant limitations in this area. In order to minimize the limitations encountered in prediction of drug-drug interaction, this study proposes a more advanced convolution neural network model than the studies in the literature, using the properties of drugs. The proposed approach is carried out in two main stages, feature extraction and design of the convolutional neural network model. In terms of results obtained with performance evaluation procedures, it is clear that the success of the proposed approach is superior to other approaches for prediction of drug-drug interaction. [ABSTRACT FROM AUTHOR]

As a kind of optical remote sensing image with high spectral resolution, hyperspectral images are 3D image cubes with rich spectral feature information. At present, hyperspectral image classification is one of the most active research areas in the remote sensing field. In recent years, deep learning and especially convolutional neural networks (CNN) have been used frequently in this field. In this study, a new model has been developed for hyperspectral image classification. The proposed model consists of 3D CNN, Squeeze-and-Excitation, and 2D Depthwise separable convolution (DSC). Since spectral feature information cannot be obtained with 2D CNN, 3D CNN is used to obtain both spectral and spatial features. Using only 3D CNN is that it increases computational complexity. The use of hybrid CNN solves this problem. In order to extract more spatial features in hybrid CNNs, 2D CNN is used after 3D CNN. However, within the scope of this study, the use of 2D DSC was preferred instead of 2D CNN. By using 2D DSC instead of standard 2D CNN, the number of trainable parameters and computational cost are reduced. This has improved the classification performance. In addition, the hybrid CNN model is combined with a Squeeze-and-Excitation block to improve performance in feature extraction and hyperspectral image classification. HyRank-Loukia and Houston 2013 datasets were used to test the classification performance of the proposed method. As a result of the applications performed by taking 10% training samples, 90.86% and 99.39% overall accuracy values were obtained, respectively. In addition, the proposed method was compared with seven different methods from the literature and it was seen that the proposed method gave better results. [ABSTRACT FROM AUTHOR]

Grasshoppers damage crops and causes millions of tons of food to be destroyed every year. The development of effective and accurate locust identification systems is critical in controlling locust species and preventing food loss. In this study, 11 different plant pest grasshopper species seen in various parts of our country and the world were classified using various convolutional neural network models. The dataset used in the study was obtained by observing the Eastern and Southeastern Anatolia regions of our country. The novelity of this study is that a dataset named GHCD11 has been created for 11 different plant pest grasshopper species in our country. In addition, VGG16, VGG19, ResNet50, DenseNet121, EfficientNet, MobileNet, which are in the Keras library and are widely used in image classification, were used for the automatic classification of 11 different grasshopper species in the study. As a result of experimental studies on the GHCD11 dataset with learning transfer, remarkable classification accuracies in the range of 95% to 99% were obtained. The study is important because it not only presents a novel dataset, but also demonstrates that automatic identification and detection of plant pest grasshopper species can be done with high accuracy using convolutional neural network architectures. [ABSTRACT FROM AUTHOR]

Copyright of International Journal of Engineering Research & Development (IJERAD) is the property of International Journal of Engineering Research & Development and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Objectives:To evaluate the performance of convolutional neural network (CNN) architectures to distinguish eyes with glaucoma from normal eyes.Materials and Methods:A total of 9,950 fundus photographs of 5,388 patients from the database of Eskişehir Osmangazi University Faculty of Medicine Ophthalmology Clinic were labelled as glaucoma, glaucoma suspect, or normal by three different experienced ophthalmologists. The categorized fundus photographs were evaluated using a state-of-the-art two-dimensional CNN and compared with deep residual networks (ResNet) and very deep neural networks (VGG). The accuracy, sensitivity, and specificity of glaucoma detection with the different algorithms were evaluated using a dataset of 238 normal and 320 glaucomatous fundus photographs. For the detection of suspected glaucoma, ResNet-101 architectures were tested with a data set of 170 normal, 170 glaucoma, and 167 glaucoma-suspect fundus photographs.Results:Accuracy, sensitivity, and specificity in detecting glaucoma were 96.2%, 99.5%, and 93.7% with ResNet-50; 97.4%, 97.8%, and 97.1% with ResNet-101; 98.9%, 100%, and 98.1% with VGG-19, and 99.4%, 100%, and 99% with the 2D CNN, respectively. Accuracy, sensitivity, and specificity values in distinguishing glaucoma suspects from normal eyes were 62%, 68%, and 56% and those for differentiating glaucoma from suspected glaucoma were 92%, 81%, and 97%, respectively. While 55 photographs could be evaluated in 2 seconds with CNN, a clinician spent an average of 24.2 seconds to evaluate a single photograph.Conclusion:An appropriately designed and trained CNN was able to distinguish glaucoma with high accuracy even with a small number of fundus photographs.

Gold is one of the most counterfeited precious metals. The color of copper is like gold. For this reason, copper is one of the most used materials for color counterfeiting. When the chemical properties are concerned, wolfram is like gold (density of gold and tungsten are 19.30 g/ml and 19.25 g/ml, respectively), so it can be used as a chemical counterfeit. The purity of gold can be determined by X-ray, but this method is costly. The current low-cost methods of jewelers have been experimented with for counterfeit gold detection in this paper. When a gold matter is hit by a subject, the sound frequency is higher than the frequency of sound when the same experiment is done with copper. Furthermore, counterfeit gold color is brighter than real ones. The color of gold is unique, and it is called "gold yellow". In this research, by employing sound and image processing, counterfeit and original gold are differentiated. For the image processing part, first a Convolutional Neural Network (CNN)-based toolbox for segmenting the gold material is applied. Then, deep CNNs for differentiating the color of the gold and copper materials are employed. Promising results are achieved with both sound and image processing techniques.

Solunum sistemine etki eden ve ileri vakalarda ölüme neden olan korona virüs salgını yaklaşık iki yıldır devam etmektedir. Her ülkenin salgın ile mücadele yöntemi farklı olmasına rağmen ortak izlenen metot ise hastalığın tespiti ve izolasyonudur. Tespit ve izolasyon için en kritik adım ise COVID-19 tanısının doğru ve hızlı konulmasıdır. Akciğer X-Ray görüntülerinde virüse özgü bulgulara rastlanılması, bu verilerin hastalık teşhisinde kullanılabileceğini göstermektedir. İlgili çalışmanın amacı, makine öğrenmesi yöntemleri ile COVID-19 ve diğer akciğer hastalıklarına ait X-Ray görüntülerini işleyerek çoklu sınıflandırma yapmaktır. Bu sayede kriz anında tanı koyma ve izolasyon için yardım alınacak alanında uzman olmayan personele mobil cihazlar vasıtasıyla karar aşamasında destek sağlanması hedeflenmektedir. Bu amaçla: COVID-19, Normal, Akciğer Opasitesi, Diğer Pnömoni etiketlerine ait 11,293 X-Ray görüntüsünden oluşan veri seti MobileNetV2, NASNetMobile, Xception ve DenseNet121 CNN ağları kullanılarak sınıflandırılmış ve sonuçlar karşılaştırılmıştır. En başarılı sonuçlar DenseNet121 ve MobileNet ağları ile elde edilmiş olup sırası ile %92,16 ve %91,78 doğruluk oranıyla sınıflandırma gerçekleştirilmiştir.

Surface defect detection is very important in manufacturing systems to ensure high quality products. Unlike manual inspections under human supervision, automatic surface defect detection is both efficient and highly accurate. In this study, a Result Weighting-based Resnet Feature Pyramid Network (SA-RÖPA) model has been developed for automatic pixel-level surface defect detection. In the first stage of the proposed model, the pre-trained Resnet50 network was used, and feature maps were extracted from the different levels of this network. In the second stage, Feature Pyramid Model was applied to these feature maps in order to hierarchically share important information in defect detection. In the third stage, 4 different error detection results were obtained by using these feature maps. In the last stage, four different results obtained using the developed Result Weighting (SA) module were effectively combined. The proposed SA-ROPA model has been tested with MT, MVTex-Doku, and AITEX datasets, which are widely used in defect detection studies. In experimental studies, the mIoU value obtained for the MT and AITEX datasets using the proposed model was calculated as 79.92%, 76.37%, and 82.72%, respectively. These results have shown that the proposed SA- ROPA model is more successful than other state-of-the-art models.

Nowadays, the use of the human ear images gains importance for the sustainability of biometric authorization and surveillance systems. Contemporary studies show that such processes can be done semi-automatically or fully automatically, instead of being done manually. Due to the fact that deep learning uses abstract features (i.e., representation learning), it reaches quite high performance values compared to classical methods. In our study, a synergistic gender recognition approach based on hybrid deep learning was created based on the use of human ear images in classifying people fully automatically according to their gender. By means of hybridization, hybrid deep neural network architectural models are used, which include both convolutional neural network component and recurrent neural network type components together. In these models, long-short term memory and gated recurrent unit are taken as recurrent neural network type components. Thanks to these components, the hybrid model extracts the relational dependencies between the pixel regions in the image very well. On account of this synergistic approach, the gender classification accuracy of hybrid models is higher than the standalone convolutional neural network model in our study. Two different image datasets with gender marking were used in our experiments. The reliability of the experimental results has been proven by objective metrics. In the conducted experiments, the highest values in gender recognition with hybrid models were obtained with the test accuracy of 85.16% for the EarVN dataset and 87.61% for the WPUT dataset, respectively. Discussion and conclusions are included in the last section of our study. [ABSTRACT FROM AUTHOR]

The subject of automatic detection of the presence of cervical cancer by evaluating histopathological Pap-Smear images with computerized diagnosis / detection software is an active field of study. The reason for this is that the objects (cell nucleus, cytoplasm, white blood cell, bacilli and speckle) in the visuals overlap and change the geometric structure and pattern of each other, they are dispersed in the image with different density, and the noise patterns are different. In addition, the difficulty and costs of creating a tagged large dataset prevented the emergence of a common dataset in this area. The mentioned difficulties negatively affect the achievements in current classification studies and trigger the need for new approaches. In this paper, a three-step approach based on building large Pap-Smear datasets using Generative Adversarial Networks (GANs) is proposed. Accordingly, in the first step, geometric shape and pattern models of each object structure in Pap-Smear images are created. In the second stage, synthetic Pap-Smear images (Ground True) with the desired number and distribution of objects are produced using the produced parametric models. In the third stage, the performances of existing GANs (Pix2Pix, CycleGAN, DiscoGAN and AttentionGAN) to produce GT are evaluated and a solution-oriented new current GAN architecture (Pix2PixSSIM) is proposed. Experimental studies show that a large Pap-Smear data set can be produced in a very short time with the proposed GAN architecture. In this way, it is seen that deep networks with high classification success can be trained. [ABSTRACT FROM AUTHOR]

Lung cancer is a common health problem in our country and in the world, and it ranks first in cancer-related deaths. Early diagnosis of lung cancer is of vital importance in terms of more informed progress about the course of the disease and survival of the patient. Recently, with the development of technology, artificial intelligence and deep learning-based systems; by using data obtained from medical imaging systems such as Computed Tomography (CT), Magnetic Resonance (MR), it provides some convenience to experts to diagnose the disease. In this study, a new Convolutional Neural Network (CNN) model is proposed to classify cancerous and normal lung CT images. The classification results of the proposed ESA model and the ResNeXt deep learning model, which is more successful than other pre-trained deep learning models in the literature, were compared. As a result of the study, it has been observed that the proposed ESA model performs better than the ResNeXt architecture, with an accuracy of 99%. In addition, the images in the current study were classified in their raw form without using any feature extraction method. And it has been observed that the proposed ESA model has a lower number of layers compared to the methods used in similar studies in the literature, as well as a higher classification success. [ABSTRACT FROM AUTHOR]

This research investigates how eye state detection is used to overcome simple problems including blinking, eyestrain, and driving drowsiness. The study utilized the ZJU dataset, and it suggested an eye state recognition approach based on image preprocessing methods and a deep learning-based convolutional neural network (CNN). First, multiple pooling layers were tested in the suggested CNN model, and it was discovered that the average pooling performed the best in the results. The image preprocessing methods applied on the ZJU dataset were then trained on the proposed CNN model and their results have been compared. The CNN model performed exceptionally well on the ZJU dataset, according to a comparison of the results obtained using the histogram equalization method, with 94.32% accuracy, 94.95% sensitivity, 92.42% specificity, 97.41% precision, and 96.16% F1 score performance metrics. The results of this investigation were compared to prior studies on the ZJU dataset, which had provided performance measures. When compared to the literature, it was obtained that the proposed technique has a high classification performance in detecting vision problems. [ABSTRACT FROM AUTHOR]

Turkey has 40% of the world's marble potential with its marble reserves in more than 120 different colors and designs. In addition, it is a country with a high potential with nearly 20 granite reserves. The classification of stones such as marble and granite is quite difficult due to the similar color and pattern. This situation poses a major problem for sector employees and customers. In order to solve this problem, a method has been proposed to classify marble and granite types in Turkey. Data sets with four mobile phones and two cameras have been created in quarries and retail stores in ten cities. Lens and imaging technology diversity was provided with 6 cameras used. In addition to stone images with ideal and non-ideal pattern-color tones, images of stones exposed to improper storage conditions and photographic errors such as light and shadow were also included in the data set. As a result of the study, a new data set containing 24480 images of 102 stones in total, 88 of which are marble and 14 of which are granite, was created. The number of images was increased to 171360 with three different data augmentation methods. Stratified Shuffle Split, K fold cross validation methods were used to prevent memorization in training and to increase the performance and reliability of networks. With the data set, 7 pre-trained convolutional neural networks were trained using transfer learning method. The most successful network is ResNet-50, with a classification success rate of 97.4%. [ABSTRACT FROM AUTHOR]

Character recognition in natural images is a very difficult problem due to many factors such as variability of light, background clutter, severe blur, inconsistent resolution and different scale depth. In addition to these features, distortions in characters and numbers are encountered in street view photographs with the effect of natural events. Detecting and reading house numbers from street views is a computer vision problem that falls under the category of natural scene text recognition. Convolutional neural network (CNN) model is one of the most commonly used deep learning (DL) methods in image analysis. In this study, firstly, CNN based DL method was applied to read characters from pictures that contain house numbers in their natural image. However, successful results could not be obtained, especially in cases where there are more than one house number in the image or when the depths are very variable. A new approach DDL (deep in deep learning) using two different CNN models was proposed to increase the accuracy of the DL method and also to reduce the data size created by natural images. The performance of the proposed DDL approach was compared with the performance of the DL approach using real data consisting of 17,618 images with 113 GB (gigabyte) size consisting of building images with GPS location information taken from 35 neighborhoods of Kayseri Metropolitan Municipality (KBB). Experimental results showed that the proposed DDL approach produced more accurate results and used less storage space than DL approach. [ABSTRACT FROM AUTHOR]

Emotion recognition from facial expression is a current research topic that can be applied in the many fields of computer vision, such as human-computer interaction, emotional computing, etc. In this study, an application for emotion recognition through deep learning was made using KDEF and PICS datasets. A new model was established using the convolutional neural network (CNN), an artificial intelligence approach that involves artificial neural networks, which is one of the deep learning techniques for attribute inference. Large datasets are needed for the high performance of deep learning. There are 4900 images in the KDEF dataset and 322 images in the PICS dataset. For this purpose, primarily due to the small number of images in the PICS dataset, image iteration was applied with the data augmentation method, and the PICS dataset was increased to 4830 images. Then, the new model developed by conducting separate training on these two different datasets was tested. Seven different classes of emotion (afraid, angry, disgusted, happy, neutral, sad, surprised) were covered in each dataset in the study conducted with a new model developed based on VGGNet which is one of the CNN models. With the developed model, a high success rate was achieved by obtaining 97.44% accuracy values in the validation set of the KDEF and 98.24% accuracy values in the validation set of the PICS dataset. [ABSTRACT FROM AUTHOR]

Lung diseases are among the diseases that seriously threaten human health, and many deaths today are caused by lung diseases. Thanks to the lung sounds, important inferences can be made about lung diseases. Doctors often use the auscultation technique to evaluate patients with lung conditions. However, this technique has some drawbacks. For example, this may lead to a misdiagnosis if the doctor has not received a good medical education. In addition, since the lung sounds are nonstationary, the analysis and recognition process is complex. Therefore, the development of automatic recognition systems will help in making more precise and accurate diagnoses. Many studies based on traditional sound processing routines have been proposed to diagnose lung diseases and to assist professionals in their diagnosis. In this study, a method based on deep learning has been proposed for the classification of lung sounds. For this purpose, the Convolutional Neural Network (CNN) has been designed. In addition, experiments are carried out using different machine learning methods based on feature extraction. Experiments to evaluate the effectiveness of different methods are carried out using the ICBHI 2017 data set consisting of four classes commonly used in the literature. On average, 64.5% accuracy is obtained from the proposed method. In addition, when the results obtained from the proposed method are compared with the latest methods in the literature, it is seen that it has a better performance in terms of classification success.

Copyright of Duzce University Journal of Science & Technology is the property of Duzce University Journal of Science & Technology and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Polytechnic is the property of Journal of Polytechnic and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

In this article, an automatic system is presented to detect age-related macular degeneration, which is the cause of 8.7% of visual impairment in the world. For the diagnosis of the disease convolutional neural network, which has recently been used in feature extraction from the image, has been utilized. In the study, the convolutional neural network was trained with 28031 labeled Optical Coherence Tomography (OCT) images to classify age-related macular degeneration. The trained convolutional neural network was tested with 6900 OCT images. Xception, VGG16, Inception V3 and Resnet50 deep learning networks were used for training and testing. The trained networks were compared and the performance of the networks was evaluated for the OCT dataset. Classification of OCT images using Resnet50 network achieved 0.95, VGG16 network achieved 0.95, InceptionV3 network achieved 0.97 and Xception network achieved accuracy of 0.98. Xception network achieved F1 score of 0.96 which is highest rate of classification on the OCT dataset. [ABSTRACT FROM AUTHOR]

Copyright of Acta Infologica is the property of Acta Infologica and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Dicle University Journal of Engineering / Dicle Üniversitesi Mühendislik Dergisi is the property of Dicle Universitesi and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of International Journal of InformaticsTechnologies is the property of Institute of Informatics, Gazi University and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Sayısal resimler üzerinde yapılan çeşitli oynamaları tespit edebilmek, gelişen yazılımların karmaşıklığından ötürü oldukça zorlaşmaktadır. Bu karmaşıklığa çözüm olarak klasik müdahale tespiti yöntemlerine ek olarak son yıllarda evrişimsel sinir ağı tabanlı yöntemler geliştirilmiştir. Böylelikle çok karmaşık müdahaleleri bile tespit edebilen ağlar eğitilebilmiştir. Bu makalede, küçük boyutlarda pencere kullanarak bölgesel müdahale tespiti yapabilen klasik yöntemlerden olan, kameranın kendisine ait olan sensörlerinden elde edilen parmakizini kullanan sensör tabanlı PRNU(Photo Response Non Uniformity) yöntemi ile yeni bir yaklaşım olan evrişimsel sinir ağı(CNN) tabanlı kamera model sınıflandırıcısı yöntemi karşılaştırılmıştır. Böylelikle hangi yöntemin daha başarılı olduğu detaylıca ortaya koyulmuştur. Toplamda 26 adet kamera modeli ve bu kamera modellerinden seçilen 96 x 96’lık piksel blokları ile eğitilen CNN modeli, hem 96 hem de 128’lik pencere boyutu kullanılarak çalışan PRNU yöntemi ile kıyaslanmıştır. Bu kıyaslama sonucunda bölgesel müdahale tespiti probleminde CNN tabanlı kamera model sınıflandırıcısının PRNU yöntemine göre daha başarılı olduğu gösterilmiştir.

Recent studies in experimental psychology revealed that a developmental correspondence exists between prediction of others' action goals and motor ability to produce the same actions. Babies interact with their environment within the first year after birth and as a result of these interactions they acquire some skills and abilities. At the same time, babies interact with other people around them and gain insight into the action goals of others. This study proposes a computational model in order to examine the developmental progression of predicting the outcome of the grasping and pushing actions on a tabletop environment. In this context, some reaching trajectories were formed towards certain objects and by combining these trajectories with the images taken from a dataset, the targeted objects, with the intention of grasping or pushing, were tried to be determined. The developmental progression for grasping and pushing actions was analyzed in four stages. The results show that the computational model can establish a direct relationship between the action recognition and the ability to predict the outcome of the actions. And the prediction ability of the grasping action was developed earlier. [ABSTRACT FROM AUTHOR]

Skin cancer is common and a serious disease, which can lead to death if not treated in time. Melanoma is the rarest and most dangerous type of skin cancer. It causes the most deaths. As in all diseases, early and correct detection of skin cancer are very important. Computer Aided Diagnosis systems can help physicians and patients make better decisions. Especially, machine learning and deep learning use effectively in Computer Aided Diagnosis systems. In this study, an automatic detection system for melanoma is suggested. To illustrate the advantage of the proposed CNN model C4Net, a comprehensive experimental study has been carried out. In addition, the proposed C4Net has been compared with not only the existing deep learning algorithms such as AlexNet, GoogLeNet, ResNet and VGGNet but also conventional machine learning algorithms such as Artificial neural networks, k-Nearest neighbor algorithm and Support vector machine. In experimental studies, C4Net, which is designed as deep neural network model for melanoma detection, has obtained more classification accuracy than other methods with 96.94%. [ABSTRACT FROM AUTHOR]