Yazar "Ucuzal H." seçeneğine göre listele

Listeleniyor 1 - 2 / 2
  • Küçük Resim Yok
    Öğe
    Classification of brain tumor types by deep learning with convolutional neural network on magnetic resonance images using a developed web-based interface
    (Institute of Electrical and Electronics Engineers Inc., 2019) Ucuzal H.; Yasar S.; Colak C.
    Automated machine learning (AutoML) algorithms developed using deep learning algorithms have been the focus of interest in many studies recently. This study aims to develop a free web-based software based on deep learning that can be utilized in the diagnosis and detection of brain tumors (Glioma/Meningioma/Pituitary) on T1-weighted magnetic resonance imaging. The Keras library, which is used in Python programming language, is utilized in the construction of the deep learning algorithm in this software. The experimental results show that this software can be used for the detection and diagnosis of three types of brain tumors. This developed web-based software can be publicly available at http://biostatapps.inonu.edu.tr/BTSY/ in both English and Turkish. © 2019 IEEE.
  • Küçük Resim Yok
    Öğe
    A Web-Based Application for Identifying Objects in Images: Object Recognition Software
    (Institute of Electrical and Electronics Engineers Inc., 2019) Ucuzal H.; Balikci Cicek A.G.I.; Arslan A.G.A.K.; Colak C.
    Object recognition is a computer vision technique for identifying objects in images or videos. Object recognition is an important output of deep learning and machine learning algorithms. For this purpose, open source, free and artificial intelligence based Object Recognition Software has been developed in order to perform object recognition operation easily.In creating this web-based software, Darkflow and Tensorflow libraries are used which are based on deep learning based Python programming language and allow the design of interactive web based applications. While performing object recognition analysis in the developed software, CNN (Convolutional Neural Networks) multiple convolution layers are uncovered hidden and useful features obtained by various calculation methods. With CNN, objects are classified, objects are detected, and objects are determned by image segmentation. A pre-trained model from COCO, a large-scale object detection, partitioning and image dataset, is used to see how the web-based software work and to evaluate the analysis outputs. Object recognition analysis is applied to ten images from this data set. According to the object recognition analysis results of the ten images, the calculated accuracy rates is examined and it is found that this web based software which is developed as open source and free access gives successful estimations in object recognition.In order to see how the web-based software works and to evaluate the analysis outputs, a pre-trained model was used from COCO (Common Objects in Context) which is a large scale object detection, partitioning and image dataset. Object recognition analysis was applied to ten images from this data set. When the accuracy ratio of the ten images calculated according to the object recognition analysis result is examined, it is determined that this web based software which is developed as open source and free access gives successful predictions in object recognition.The developed software is new user-friendly web-based software that can easily identify objects in images and discriminatory from each other objects. In the following studies, in order to increase the diagnostic accuracy of the objects in the images, it is suggested that the softwares that uses deeper neural networks should be developed and the necessary infrastructure to detect the defects in the medical images can be developed. © 2019 IEEE.