Behçet Uğur Töreyin

Behçet Uğur Töreyin
Istanbul Technical University · Informatics Institute

PhD

About

143
Publications
39,811
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
2,612
Citations
Introduction
Dr. Töreyin's research interests broadly lie in signal processing and pattern recognition with applications to image/video analysis and communication systems. His research is focused on developing algorithms to analyze the content of signals from a multitude of sensors such as visible/infra-red/hyperspectral cameras, microphones, passive infra-red sensors, vibration sensors, and spectrum sensors for wireless communications. For more information, visit http://spacing.itu.edu.tr/.

Publications

Publications (143)
Preprint
Full-text available
Despite recent advances, object detection in aerial images is still a challenging task. Specific problems in aerial images makes the detection problem harder, such as small objects, densely packed objects, objects in different sizes and with different orientations. To address small object detection problem, we propose a two-stage object detection f...
Article
Myelin is an essential component of the nervous system and myelin damage causes demyelination diseases. Myelin is a sheet of oligodendrocyte membrane wrapped around the neuronal axon. In the fluorescent images, experts manually identify myelin by co-localization of oligodendrocyte and axonal membranes that fit certain shape and size criteria. Becau...
Article
Full-text available
Despite recent advances, object detection in aerial images is still a challenging task. Specific problems in aerial images makes the detection problem harder, such as small objects, densely packed objects, objects in different sizes and with different orientations. To address small object detection problem, we propose a two-stage object detection f...
Preprint
Full-text available
We propose a novel knowledge distillation methodology for compressing deep neural networks. One of the most efficient methods for knowledge distillation is hint distillation, where the student model is injected with information (hints) from several different layers of the teacher model. Although the selection of hint points can drastically alter th...
Preprint
Click-stream data, which comes with a massive volume generated by the human activities on the websites, has become a prominent feature to identify readers' characteristics by the newsrooms after the digitization of the news outlets. It is essential to have elastic architectures to process the streaming data, particularly for unprecedented traffic,...
Preprint
The recent spike in the demand for high-performance computing (HPC) server systems has created many challenges in data centers (DCs) including thermal management, system reliability sustenance and server failure minimalization. Lately, deep neural networks applied to infrared thermography (IRT) images have been successfully used for fault diagnosis...
Chapter
With the rapid increase in the number of flights all over the world, the management and control of flight operations has become difficult in recent years. Moreover, the expectations for the aviation sector indicate that this increase will continue in the upcoming years. Therefore, safer and systematic monitoring systems by eliminating the requireme...
Chapter
Human brain effectively integrates prior knowledge to new skills by transferring experience across tasks without suffering from catastrophic forgetting. In this study, to continuously learn a visual classification task sequence, we employed a neural network model with lateral connections called Progressive Neural Networks (PNN). We sparsified PNNs...
Chapter
We propose compact and effective network layer Rotational Duplicate Layer (RDLayer) that takes the place of regular convolution layer resulting up to 128\(\times \) in memory saving. Along with network accuracy, memory and power constraints affect design choices of computer vision tasks performed on resource-limited devices such as FPGAs (Field Pro...
Article
The recently popular Deep Neural Networks (DNNs) have a significant effect on the improvement of segmentation accuracy from various perspectives, including robustness and completeness in comparison to conventional methods. We figured out that the naive U-Net has some lacks in specific perspectives and there is high potential for further enhancement...
Article
Background and Objective: Visual expression of invasive breast cancer with immunohistochemistry (IHC) allows evaluation of CerbB2 receptors, such that CerbB2 mutated breast carcinomas are suitable for targeted therapy. Breast tumors are evaluated in four different scores as 0, 1, 2, 3 to decide if it is suitable for the CerbB2 protein specific trea...
Article
The cervical cancer developing from the precancerous lesions caused by the human papillomavirus (HPV) has been one of the preventable cancers with the help of periodic screening. Cervical intraepithelial neoplasia (CIN) and squamous intraepithelial lesion (SIL) are two types of grading conventions widely accepted by pathologists. On the other hand,...
Article
Full-text available
Data compression techniques allow data size to be reduced prior to data transmission and involve decompression upon transfer. This study shows for the first time that license plate (LP) detection can be accomplished without full decompression of the encoded data. Therefore, by determining in advance which images are required for LP recognition, com...
Preprint
Full-text available
Graph autoencoders are very efficient at embedding graph-based complex data sets. However, most of the autoencoders have shallow depths and their efficiency tends to decrease with the increase of layer depth. In this paper, we study the effect of adding residual connections to shallow and deep graph variational and vanilla autoencoders. We show tha...
Conference Paper
Full-text available
As cloud computing applications have witnessed an exponential growth in recent years, the maintenance of server uptime has never been more important. A fault in a server due to overload, attack, or a misconfiguration of a cooling system can be of imponderable economic and financial loss to the echelons of global institutions. The advent of thermal...
Conference Paper
Analyses of morphology, polarity, and motility of cells is important for cell biology research such as metastatic and invasive capacity of cells, wound healing, and embryonic development. Automation of such analyses using image series of phase-contrast optical microscopy, which allows label-free imaging of live cells in their living environment, is...
Preprint
Myelin is an essential component of the nervous system and myelin damage causes demyelination diseases. Myelin is a sheet of oligodendrocyte membrane wrapped around the neuronal axon. In the fluorescent images, experts manually identify myelin by co-localization of oligodendrocyte and axonal membranes that fit certain shape and size criteria. Becau...
Article
Effective compression of hyperspectral (HS) images is essential due to their large data volume. Since these images are high dimensional, processing them is also another challenging issue. In this work, an efficient lossy HS image compression method based on enhanced multivariance products representation (EMPR) is proposed. As an efficient data deco...
Conference Paper
Real-time flame detection is crucial in video-based surveillance systems. We propose a vision-based method to detect flames using Deep Convolutional Generative Adversarial Neural Networks (DCGANs). Many existing supervised learning approaches using convolutional neural networks do not take temporal information into account and require a substantial...
Conference Paper
In extracellular neural recordings, the actual signal is separated into noise and action potentials using some threshol-ding methods. Generally, the threshold is determined as 3 to 5 times the estimated standard deviation of the noise in the filtered recordings. However, the value of the standard deviation estimate in all of these methods depends o...
Article
Background: The myelin sheath produced by glial cells insulates the axons, and supports the function of the nervous system. Myelin sheath degeneration causes neurodegenerative disorders, such as multiple sclerosis (MS). There are no therapies for MS that promote remyelination. Drug discovery frequently involves screening thousands of compounds. Ho...
Conference Paper
The segmentation of cells is necessary for biologists in the morphological statistics for quantitative and qualitative analysis in Phase-contrast Microscopy (PCM) images. In this paper, We address the problem of the segmentation task in PCM images. Deep Neural Networks (DNNs) commonly initialized with weights from a network that pre-trained on a la...
Preprint
Full-text available
Human brain effectively integrates prior knowledge to new skills by transferring experience across tasks without suffering from catastrophic forgetting. In this study, to continuously learn a visual classification task sequence (PermutedMNIST), we employed a neural network model with lateral connections, sparse group LASSO regularization and projec...
Conference Paper
Full-text available
Graphs are usually represented by high dimensional data. Hence, graph embedding is an essential task, which aims to represent a graph in a lower dimension while protecting the original graph’s properties. In this paper, we propose a novel graph embedding method called Residual Variational Graph Autoencoder (RVGAE), which boosts variational graph au...
Conference Paper
Object recognition can be performed with high accuracy thanks to the robust feature descriptors defining the significant areas in images. However, these features suffer from high dimensional structure, in other words "curse of dimensionality" for further processes. Autoencoders (AE) are proposed in this study to solve the dimensionality reduction p...
Conference Paper
Myelin sheath, wrapped around axons, allows rapid neural signal transmission, and degeneration of myelin causes various neurodegenerative diseases, such as, Multiple Sclerosis (MS). For candidate drug discovery, it is essential to quantify myelin. This requires tedious expert labor comprising myelin labelling on microscopic fluorescence images, usu...
Conference Paper
Phase contrast optical microscopy is a preferred imaging technique for live-cell, temporal analysis. Segmentation of cells from time series data acquired with this technique is a labor-intensive and time-consuming task that cell biology researchers need solution for. In this study traditional image processing and deep learning based approaches for...
Conference Paper
The quantitative and qualitative ascertainment of cell culture is integral to the robust determination of the cell structure analysis. Microscopy cell analysis and the epithet structures of cells in cell cultures are momentous in the fields of the biological research process. In this paper, we addressed the problem of phase-contrast microscopy unde...
Conference Paper
Full-text available
In this study, the data which is compressed by the most recent video compression standard HEVC (High Efficient Video Coding) has been analyzed and shown that vehicle license plate can be detected without a need to full decode process. An image has been created that represents HEVC block partitioning structure which is available at the beginning of...
Article
Background: Finding analyzable metaphase chromosome images is an essential step in karyotyping which is a common task for clinicians to diagnose cancers and genetic disorders precisely. This step is tedious and time- consuming. Hence developing automated fast and reliable methods to assist clinical technicians becomes indispensable. Previous approa...
Article
A sparse and low-rank matrix decomposition-based method is proposed for anomaly detection in hyperspectral data. High-dimensional data are decomposed into low-rank and sparse matrices representing background and anomalies, respectively. The problem of the decomposition process is defined from the dictionary learning point of view. Therefore, our wa...
Preprint
Full-text available
Real-time flame detection is crucial in video based surveillance systems. We propose a vision-based method to detect flames using Deep Convolutional Generative Adversarial Neural Networks (DCGANs). Many existing supervised learning approaches using convolutional neural networks do not take temporal information into account and require substantial a...
Article
The fish passage performance and flow structure of a brush fish pass were investigated at the ̇Incirli SmallHydropower Plant on the ̇Iyidere River, located in the East Black Sea region of Turkey. The spatial distributions of velocityvectors, power velocity, Froude number and turbulent kinetic energy are presented. The flow is quasi-uniform andsubcr...
Preprint
Full-text available
The cervical cancer developing from the precancerous lesions caused by the Human Papilloma Virus (HPV) has been one of the preventable cancers with the help of periodic screening. There are two types of grading conventions widely accepted among pathologists. On the other hand, inter-observer variability is an important issue for final diagnosis. In...
Article
Full-text available
The pages that appear in front of users on digital platforms used for online advertising to attract attention to target product are called landing pages. Landing pages aim to increase advertisement conversion rate using the metrics like clicks, views or subscribes. In this study, a method is presented to automatically classifier the most commonly u...
Preprint
Full-text available
Oligodendrocytes wrap around the axons and form the myelin. Myelin facilitates rapid neural signal transmission. Any damage to myelin disrupts neuronal communication leading to neurological diseases such as multiple sclerosis (MS). There is no cure for MS. This is, in part, due to lack of an efficient method for myelin quantification during drug sc...
Article
Hyperspectral imaging features an important issue in remote sens ing and applications. Requirement to collect high volumes of hyper spectral data in remote sensing algorithms poses a compression prob lem. To this end, many techniques or algorithms have been develop ed and continues to be improved in scientific literature. In this paper, we propose...
Article
Full-text available
Vehicle logo recognition has become an important part of object recognition in recent years because of its usage in surveillance applications. In order to achieve a higher recognition rates, several methods are proposed, such as Scale Invariant Feature Transform (SIFT), convolutional neural networks, bag-of-words and their variations. A fast logo r...
Conference Paper
Full-text available
This study proposes a unique approach to classify CerbB2 tumor cell scores in breast cancer based on deep learning models. Another contribution of the study is the creation of a dataset from original breast cancer tissues. On the purpose of training, validating and testing with deep learning models cell fragments were generated from sample tissue i...
Conference Paper
Cervical carcinoma is one of the frequently seen cancers in the world and in our country, develops from precursor lesions. These precursor lesions are analyzed by pathologists so that the diagnosis of the disease can be made. In this study, a system that performs automatic detection of pre-cancerous lesions was performed using the convolutional neu...
Chapter
In this chapter, a pyroelectric infrared (PIR) sensor-based flame detection system is described using a Markovian decision algorithm. The chapter is based on the paper in [ 1 ]. A differential PIR sensor is only sensitive to sudden temperature variations within its viewing range and it produces a time-varying signal. The wavelet transform of the PI...
Chapter
To avoid large-scale fire and smoke damage, timely and accurate fire detection is essential. The sooner the fire is detected, the better the chances are for survival. However, not only is early detection crucial, but it is also important to have a clear understanding of the fire development and the location. Where did the fire start? What is the si...
Article
Full-text available
In this paper, an online learning framework called adaptive decision fusion (ADF) is employed for short-term wind speed and turbulence intensity forecasting by use of wind speed data for the city İzmit located in the North West of Turkey for each season. Fixed-weight linear combination (FW) is derived and used to stand comparison with ADF. Wind spe...
Conference Paper
Hyperspectral data is composed of a set of correlated band images. In order to efficiently compress the hyperspectral imagery, this inherent correlation may be exploited by means of spectral decorrelators. In this paper, a fractional wavelet transform based method is introduced for spectral decorrelation of hyperspectral data. As opposed to regular...
Conference Paper
Full-text available
Fluorescence in situ hybridization (FISH) technique widely used in cancer diagnosis is based on displaying chromosomal regions as FISH signals by staining with specific dyes. In this study, a new multi-level thresholding based FISH signal segmentation method is proposed for images produced by FISH technique. Cell nuclei are segmented on images, tha...
Conference Paper
Sparse coding based compression of hyperspectral imagery yields better rate-distortion performance especially for low bit-rates when compared with other state-of-the-art methods in the literature. In this paper, an on-line dictionary learning based lossy compression method is proposed yielding even a better rate-distortion performance, thanks to th...
Conference Paper
In preceding paper, a compression algorithm for hyperspectral images using a novel multivariate data decomposition method called Enhanced Multivariance Products Representation (EMPR) is developed. The test results obtained by performing some EMPR approximations of different orders and their qualities are reported. In order to improve performance, E...
Article
The joint estimation of the parameters and the states of the hemodynamic model from the blood oxygen level dependent (BOLD) signal is a challenging problem. In the functional magnetic resonance imaging (fMRI) literature, quite interestingly, many proposed algorithms work only as a filtering method. This makes the estimation of hidden states and par...
Book
This book describes the signal, image and video processing methods and techniques for fire detection and provides a thorough and practical overview of this important subject, as a number of new methods are emerging. This book will serve as a reference for signal processing and computer vision, focusing on fire detection and methods for volume senso...
Chapter
In the last decades, huge improvements in the computational power of computers and decreasing cost of imaging sensors have made it possible to use video-based fire detection techniques in real-time applications. Most video fire detection algorithms use visible range cameras because they are much cheaper than infrared (thermal) and time-of-flight ca...
Article
Full-text available
Sparse models provide data representations in the fewest possible number of nonzero elements. This inherent characteristic enables sparse models to be utilized for data compression purposes. Hyperspectral data is large in size. In this paper, a framework for sparsity-based hyperspectral image compression methods using online learning is proposed. T...
Article
Hyperspectral images have huge data volume that contains spectral and spatial information. This high data volume leads to processing, storage, and transmission problems. Moreover, insufficient training data results in Hughes phenomenon. It is possible to solve these problems with the help of feature selection. In this paper, a method that evaluates...
Article
Full-text available
In this paper, to decrease the computational cost and number of cycles in Template Matching Algorithm, a novel two-stage algorithm is proposed. The Sum of Absolute Differences method is used for matching. The proposed algorithm is implemented on Field-Programmable-Gate-Array (FPGA). The algorithm is accelerated with the effective usage of Block RAM...
Conference Paper
Full-text available
Hyperspectral remote sensing (HSRS) is becoming more and more attractive. Recent advances in sensor technologies enabled numerous applications of this imaging modality. HSRS research has been conducted at TUBITAK UZAY since 2012. This paper provides a short survey of these research and development activities ranging from hyperspectral remote sensin...
Conference Paper
Integer-coefficient Discrete Wavelet Transformation (DWT) filters widely used in the literature are implemented and investigated as spectral decorrelator. As the performance of spectral decorrelation step has direct impact on the compression ratio (CR), it is important to employ the most convenient spectral decorrelator in terms of computational co...
Article
A hyperspectral image compression method is proposed using an online dictionary learning approach. The online learning mechanism is aimed at utilizing least number of dictionary elements for each hyperspectral image under consideration. In order to meet this "sparsity constraint", basis pursuit algorithm is used. Hyperspectral imagery from AVIRIS d...
Article
Full-text available
Sparse coding ensures to express the data in terms of a few nonzero dictionary elements. Since the data size is large for hyperspectral imagery, it is reasonable to use sparse coding for compression of hyperspectral images. In this paper, a hyperspectral image compression method is proposed using a discriminative online learning-based sparse coding...