Browsing by Author "Karaci A."

Now showing 1 - 14 of 14

Computer Control with Face and Eye Movements Using Deep Learning and Image Processing Methods
(2022-12-31) Çapşek M.F.; Karaci A.
In this study, an artificial intelligence-supported system was developed for these individuals to control the mouse with head and eye movements. In this system, facial movements and eyes are detected in real-time from the images obtained from the camera through the Haar Cascade, Dlib, and Open CV libraries. While Haar Cascade is used to detect the face region, the Dblib library is used to obtain right and left eye region images from this detected face image. These eye region images obtained are given as input to the CNN model trained with 2874 eye data and it is determined whether the eye is closed or open. The CNN model was trained on a public eye image dataset representing 1500 open and 1374 closed eye states. Closing and opening of the left eye cause the left click of the mouse and closing and opening of the right eye cause the right click of the mouse. In addition, the position of the face detected by Haar Cascade is used to model mouse movement. According to the test results, it has been observed that the system correctly detects the eyes and the open-closed state of these eyes, and correctly classifies the blinking event in both eyes with CNN. However, it has been determined that there are cases of slowness or incomplete adaptation to facial movement in the modeling of mouse movement.
Detection and classification of shoulder implants from X-ray images: YOLO and pretrained convolution neural network based approach
(2022-01-01) Karaci A.; Karaci, A
Shoulder implants may need to be replaced several months or years after insertion. In this case, it is important to determine the manufacturer or model of the implant. In some cases, the implant manufacturer and model may not be known to patients or their physicians due to uncertainty in medical records. Today, the task of identifying an implant manufacturer or model in such situations relies on meticulous examination and visual comparison of X-ray images taken from the implant by medical professionals. But this identification task is often time-consuming, error-prone and difficult for both radiologists and orthopedic surgeons. In this study, it is aimed to automatically detect the implant manufacturer using deep learning methods. For this purpose, pretrained CNN architectures and cascade models consisting of feeding these architectures with the YOLO algorithm have been proposed.
Detection and classification of shoulder implants from X-ray images: YOLO and pretrained convolution neural network based approach
(2022-01-01) Karaci A.
Shoulder implants may need to be replaced several months or years after insertion. In this case, it is important to determine the manufacturer or model of the implant. In some cases, the implant manufacturer and model may not be known to patients or their physicians due to uncertainty in medical records. Today, the task of identifying an implant manufacturer or model in such situations relies on meticulous examination and visual comparison of X-ray images taken from the implant by medical professionals. But this identification task is often time-consuming, error-prone and difficult for both radiologists and orthopedic surgeons. In this study, it is aimed to automatically detect the implant manufacturer using deep learning methods. For this purpose, pretrained CNN architectures and cascade models consisting of feeding these architectures with the YOLO algorithm have been proposed.
Determining students' level of page viewing in intelligent tutorial systems with artificial neural network
(2014-03-01) Karaci A.; Arici N.
The concept of level of page viewing (LPV) refers to the extent to which a student actively revises the pages that he or she has to study in tutorial systems. In the present study, an artificial neural network (ANN) model, which is composed of 5 inputs, 20 and 30 neurons, 2 hidden layers, and 1 output, was designed to determine the students' LPV. After this network was trained, it was integrated into a web-based prototype teaching system, which was developed by ASP.net C# programming language. Additionally, Decision Tree method is tried to determine students' LPV. However, this method gave wrong results according to expected LPV values. In this system, the student first studies the pages uploaded by the teacher onto the system. After studying all the pages within the scope of a topic, the student can go to the test page for evaluation purposes. LPVs of a student who wants to navigate to the test page are calculated by an ANN module added to the system. On the condition that one or more of the LPV's are not up to the desired level, the student is not allowed to take the test and is informed of the pages with missing LPV's so that he can re-study these pages. This prototype system developed based on ANN to determine students' LPV is essential for intelligent tutorial systems, geared to provide intelligent assistance and guidance. The system can track the pages which the students did not study sufficiently and thus direct them to relevant pages. How much activity the students perform on each page to study is observed before they actually take the test, and the areas which should be further revised are determined much in advance. © 2012 Springer-Verlag London.
Estimating the properties of ground-waste-brick mortars using DNN and ANN
(2019-01-01) Karaci A.; Yaprak H.; Ozkaraca O.; Demir I.; Simsek O.
In this study, deep-neural-network (DNN)- and artificial-neural-network (ANN)-based models along with regression models have been developed to estimate the pressure, bending and elongation values of ground-brick (GB)-added mortar samples. This study is aimed at utilizing GB as a mineral additive in concrete in the ratios 0.0%, 2.5%, 5.0%, 7.5%, 10.0%, 12.5% and 15.0%. In this study, 756 mortar samples were produced for 84 different series and were cured in tap water (W), 5% sodium sulphate solution (SS5) and 5% ammonium nitrate solution (AN5) for 7 days, 28 days, 90 days and 180 days. The developed DNN models have three inputs and two hidden layers with 20 neurons and one output, whereas the ANN models have three inputs, one output and one hidden layer with 15 neurons. Twenty-five previously obtained experimental sample datasets were used to train these developed models and to generate the regression equation. Fifty-nine non-training-attributed datasets were used to test the models. When these test values were attributed to the trained DNN, ANN and regression models, the brick-dust pressure as well as the bending and elongation values have been observed to be very close to the experimental values. Although only a small fraction (30%) of the experimental data were used for training, both the models performed the estimation process at a level that was in accordance with the opinions of experts. The fact that this success has been achieved using very little training data shows that the models have been appropriately designed. In addition, the DNN models exhibited better performance as compared with that exhibited by the ANN models. The regression model is a model whose performance is worst and unacceptable; further, the prediction error is observed to be considerably high. In conclusion, ANN- and DNN-based models are practical and effective to estimate these values.
Expansin gene family database: A comprehensive bioinformatics resource for plant expansin multigene family
(2023-06-01) Kök B.Ö.; Celik Altunoglu Y.; Öncül A.B.; Karaci A.; Cengiz Baloglu M.
Expansins, which are plant cell wall loosening proteins associated with cell growth, have been identified as a multigene family. Plant expansin proteins are an important family that functions in cell growth and many of developmental processes including wall relaxation, fruit softening, abscission, seed germination, mycorrhiza and root nodule formation, biotic and abiotic stress resistance, invasion of pollen tube stigma and organogenesis. In addition, it is thought that increasing the efficiency of plant expansin genes in plants plays a significant role, especially in the production of secondary bioethanol. When the studies on the expansin genes are examined, it is seen that the expansin genes are a significant gene family in the cell wall expansion mechanism. Therefore, understanding the efficacy of expansin genes is of great importance. Considering the importance of this multigene family, we aimed to create a comprehensively informed database of plant expansin proteins and their properties. The expansin gene family database provides comprehensive online data for the expansin gene family members in the plants. We have designed a new website accessible to the public, including expansin gene family members in 70 plants and their features including gene, coding and peptide sequences, chromosomal location, amino acid length, molecular weight, stability, conserved motif and domain structure and predicted three-dimensional architecture. Furthermore, a deep learning system was developed to detect unknown genes belonging to the expansin gene family. In addition, we provided the blast process within the website by establishing a connection to the NCBI BLAST site in the tools section. Thus, the expansin gene family database becomes a useful database for researchers that enables access to all datasets simultaneously with its user-friendly interface. Our server can be reached freely at the following link (http://www.expansingenefamily.com/).
Intelligent tutoring system model based on fuzzy logic and constraint-based student model
(2019-08-01) Karaci A.
A model for an intelligent tutoring system (ITS) that uses fuzzy logic and a constraint-based student model (CBM) is proposed. The goal of the ITS is to teach the use of punctuation in Turkish. The proposed ITS includes two student models, i.e., an overlay student model and a CBM. The student modeler in the CBM records each mistake a student make when answering questions in the system. Immediate feedback and hints are provided based on the recorded mistakes. In addition, moreover the level of students’ learning of the usage of punctuation marks is determined and overlay student model is updated according to the mistakes. If the student cannot provide the correct answer relative to the desired learning level after a specified number of attempts, this information is recorded by the overlay student model. Students can study the pages and attempt to answer the questions again. For determining the level of learning MYCIN certainty factor, the number of times the student takes for answering the question and fuzzy logic decision system are used. Crowded classes make it difficult for teachers to evaluate all student answers and provide individual feedback. The proposed ITS identifies student mistakes and provides feedback immediately. In addition, the ITS analyzes mistakes to determine the student’s learning gaps relative to specific topics and concepts. Learning to use punctuation correctly is valuable; thus, the proposed ITS model is important and worthwhile.
Intelligent tutoring system model based on fuzzy logic and constraint-based student model
(2019-08-01) Karaci A.; Karaci, A
A model for an intelligent tutoring system (ITS) that uses fuzzy logic and a constraint-based student model (CBM) is proposed. The goal of the ITS is to teach the use of punctuation in Turkish. The proposed ITS includes two student models, i.e., an overlay student model and a CBM. The student modeler in the CBM records each mistake a student make when answering questions in the system. Immediate feedback and hints are provided based on the recorded mistakes. In addition, moreover the level of students’ learning of the usage of punctuation marks is determined and overlay student model is updated according to the mistakes. If the student cannot provide the correct answer relative to the desired learning level after a specified number of attempts, this information is recorded by the overlay student model. Students can study the pages and attempt to answer the questions again. For determining the level of learning MYCIN certainty factor, the number of times the student takes for answering the question and fuzzy logic decision system are used. Crowded classes make it difficult for teachers to evaluate all student answers and provide individual feedback. The proposed ITS identifies student mistakes and provides feedback immediately. In addition, the ITS analyzes mistakes to determine the student’s learning gaps relative to specific topics and concepts. Learning to use punctuation correctly is valuable; thus, the proposed ITS model is important and worthwhile.
Predicting Breast Cancer with Deep Neural Networks
(2020-01-01) Karaci A.; Karaci, A
In this study, a deep neural network (DNN) MODEL was developed which diagnoses breast cancer using information about age, BMI, glucose, insulin, homa, leptin, adiponectin, resistin and MCP-1. The data used in this model was collected by Patrício et al. [7] from 116 women of which 64 has breast cancer and 52 do not. While 70% of this data (81 cases) was used for instructing the DNN model, 30% (35 cases) was used for testing. The DNN model was created in Python programming language using Keras Deep Learning Library. After model creation, machine learning was conducted using probable optimisation algorithms, loss functions and activation functions and the best three models were saved. For performance evaluation of the models, metrics of specificity, sensitivity and accuracy were employed. The specificity values of the best three models were calculated as [0.882, 0.941] and sensitivity values were found to be [0.888, 0.944]. In other words, while the models predict healthy women at the rates of minimum 88.2% and maximum 94.1%; they predict women with breast cancer at the rates of minimum 88.8% and 94.4%. For both women with and without breast cancer these prediction rates are sufficient and much higher than those reported by Patrício et al. [7].
Predicting Breast Cancer with Deep Neural Networks
(2020-01-01) Karaci A.
In this study, a deep neural network (DNN) MODEL was developed which diagnoses breast cancer using information about age, BMI, glucose, insulin, homa, leptin, adiponectin, resistin and MCP-1. The data used in this model was collected by Patrício et al. [7] from 116 women of which 64 has breast cancer and 52 do not. While 70% of this data (81 cases) was used for instructing the DNN model, 30% (35 cases) was used for testing. The DNN model was created in Python programming language using Keras Deep Learning Library. After model creation, machine learning was conducted using probable optimisation algorithms, loss functions and activation functions and the best three models were saved. For performance evaluation of the models, metrics of specificity, sensitivity and accuracy were employed. The specificity values of the best three models were calculated as [0.882, 0.941] and sensitivity values were found to be [0.888, 0.944]. In other words, while the models predict healthy women at the rates of minimum 88.2% and maximum 94.1%; they predict women with breast cancer at the rates of minimum 88.8% and 94.4%. For both women with and without breast cancer these prediction rates are sufficient and much higher than those reported by Patrício et al. [7].
Prediction of the effect of varying cure conditions and w/c ratio on the compressive strength of concrete using artificial neural networks
(2013-01-01) Yaprak H.; Karaci A.; Demir I.
The present study aims at developing an artificial neural network (ANN) to predict the compressive strength of concrete. A data set containing a total of 72 concrete samples was used in the study. The following constituted the concrete mixture parameters: two distinct w/c ratios (0.63 and 0.70), three different types of cements and three different cure conditions. Measurement of compressive strengths was performed at 3, 7, 28 and 90 days. Two different ANN models were developed, one with 4 input and 1 output layers, 9 neurons and 1 hidden layer, and the other with 5, 6 neurons, 2 hidden layers. For the training of the developed models, 60 experimental data sets obtained prior to the process were used. The 12 experimental data not used in the training stage were utilized to test ANN models. The researchers have reached the conclusion that ANN provides a good alternative to the existing compressive strength prediction methods, where different cements, ages and cure conditions were used as input parameters. © 2011 Springer-Verlag London Limited.
Prediction of traumatic pathology by classifying thorax trauma using a hybrid method for emergency services
(2020-12-18) Karaci A.; Ozkaraca O.; Acar E.; Demir A.
In recent years, data mining and algorithm-based methods have been used frequently for the prediction and diagnosis of various diseases. Traumas, being one of the significant health problems in the world, are also one of the most important causes of death. This study aims to predict the presence of traumatic pathology in the lung of the patients admitted to the emergency department due to blunt thorax trauma with no X-ray and computed tomography (CT) history by machine learning methods. The models developed in the study using the 5-fold cross-validation method are most accurately classified by the ensemble (voting) classifier, whether there is a pathology in X-ray (mean accuracy = 0.82) and CT (mean accuracy = 0.83). The K-nearest neighbourhood method classifies patients with pathology in X-ray by 83% accuracy, while the ensemble (voting) method classifies non-pathology patients by 94% accuracy in models. Of CT results, random forest, ensemble (voting), and ensemble (stacking) classifiers are precisely classified by 96%, while those patients with pathology are classified perspicuously by 77%. As a result, a mathematical framework using data mining methods was proposed based on estimating the X-ray and CT results for the thorax graph scan.
The prediction of potential energy and matter production from biomass pyrolysis with artificial neural network
(2017-11-01) Aydinli B.; Caglar A.; Pekol S.; Karaci A.
The potentiality determination of renewable energy resources is very important. The biomass is one of the alternative energy and material resources. There is great effort in their conversion to precious material but yet there is no generalized rule. Therefore, the prediction of the energy and material potentials of these resources has gained great importance. Also, the solution to environmental problems in real time can be found easily by predicting models. Here, the basic products of pyrolysis process, char, tar and gas were also predicted by artificial neural network modelling. The half of data obtained from real experimental process along with some content and proximate analysis were fed into artificial neural network modelling. After the training of the model with this data, the remaining half of the data were introduced into this artificial neural network model. And the model predicted the pyrolysis process products (char, tar and gaseous material). The predicted data and the real experimental data were compared. In addition, another aim of this study is to reduce the labour in identification and characterization of the pyrolysis products. For this purpose, a theoretical framework has also been sketched. The necessity of a generalized rule for generation of energy and matter production from biomass pyrolysis has been punctuated. As a result, the ANN modelling is found to be applicable in the prediction of pyrolysis process. Also, the extensive reduction in labour and saving in economy is possible.
The pyrolysis process verification of hydrogen rich gas (H-rG) production by artificial neural network (ANN)
(2016-03-02) Karaci A.; Caglar A.; Aydinli B.; Pekol S.
The main aim of this study is subject of thermochemical conversion process data into computational modelling. Especially, prediction of hydrogen gas from the pyrolysis of waste materials regarded as environmentally pollutants were accomplished by Artificial Neural Network (ANN) in context of sustainability. The data obtained from pyrolysis of biomass wastes; cotton cocoon shell (cotton-S), tea waste (tea-W) and olive husk (olive-H) were categorized and hydrogen rich gas (H-rG) portion was introduced to the NFTOOL of MATLAB program for ANN. The variables in the pyrolysis process were catalyst type, amount, temperature and biomass diversity. The H-rG production was rendered by catalysts; ZnCl2, NaCO3 and K2CO3. The combination of following condition; ZnCl2-10%, Olive-H and 973 K yield the best ANN models. This helped us save comprehensive amount of labour and time during experimentations, which also result in sharpness data in energy and environmental issues and were very ambiguous. The results were discussed by using new concepts related with energy resources, hydrogen gas, modelling and sustainability. The presented perspective here can be a useful tool for researchers and users as well as planners.