Bilgisayar Mühendisliği Bölümü Koleksiyonu

Permanent URI for this collectionhttps://hdl.handle.net/20.500.11779/1940

Browse

Now showing 1 - 20 of 142

Citation - WoS: 8
Citation - Scopus: 5
Recognizing Non-Manual Signs in Turkish Sign Language
(IEEE, 2019) Gökberk, Berk; Akarun, Lale; Aktaş, Müjde
Recognition of non-manual components in sign language has been a neglected topic, partly due to the absence of annotated non-manual sign datasets. We have collected a dataset of videos with non-manual signs, displaying facial expressions and head movements and prepared frame-level annotations. In this paper, we present the Turkish Sign Language (TSL) non-manual signs dataset and provide a baseline system for non-manual sign recognition. A deep learning based recognition system is proposed, in which the pre-trained ResNet Convolutional Neural Network (CNN) is employed to recognize question, negation side to side and negation up-down, affirmation and pain movements and expressions. Our subject independent method achieves 78.49% overall frame-level accuracy on 483 TSL videos performed by six subjects, who are native TSL signers. Prediction results of consecutive frames are filtered for analyzing the qualitative results.
Citation - Scopus: 4
Multi-View Reconstruction of 3d Human Pose With Procrustes Analysis
(IEEE, 2019) Gökberk, Berk; Akarun, Lale; Temiz, Hüseyin
Recovery of 3D human pose from cameras has been the subject of intensive research in the last decade. Algorithms that can estimate the 3D pose from a single image have been developed. At the same time, many camera environments have an array of cameras. In this paper, after aligning the poses obtained from single images using Procrustes Analysis, median filtering is utilized to eliminate outliers to find final reconstructed 3D body joint coordinates. Experiments performed on the CMU Panoptic, and Human3.6M databases demonstrate that the proposed system achieves accurate 3D body joint reconstructions. Additionally, we observe that camera selection is useful to decrease the system complexity while attaining the same level of reconstruction performance.
The Neural Correlates of the Effect of Belief in Free Will on Third-Party Punishment: an Optical Brain Imaging (fnirs) Study
(Cognitive Science Society, 2022) Çakar, Tuna; Akyürek, Güçlü; Erözden, Ozan; Şahin, Türkay; Keskin, İrem Nur; Ünlü, Meryem; Özen, Deniz Hazal; Özen, Zeynep
Third party punishment (TPP), or altruistic punishment, is specifically human prosocial behavior. TPP denotes the administration of a sanction to a transgressor by an individual that is not affected by the transgression. In some evolutionary accounts, TPP is considered crucial for the stability of cooperation and solidarity in larger groups formed by genetically unrelated individuals. Belief in free will (BFW), on the other hand, is the idea that humans have control over their behavior. BFW is a human universal notion that, in some studies, has been found to be supportive of prosocial behavior. In our study, we examined the effect of BFW on TPP under high and low affect scenarios through optical brain imaging (fNIRS). We hypothesized that in low affect cases, there would be a positive correlation between the strength of the BFW and the severity of the punishment inflicted. Obtained results and related statistical analyses indicate that participants with higher degree of BFW have more neural activation in their right dorsolateral prefrontal cortex (DLPFC) (hbo and hbt measures) in high affect scenarios, whereas the participants with lower degree of BFW have higher levels of neural activation in the medial PFC (hbo and hbt measures) in low affect scenarios. These empirical findings are in line with the research findings in the relevant academic literature and support the hypothesis that the degree of BFW influences punishment decisions.
An Exploratory Study on the Effect of Contour Types on Decision Making Via Optic Brain Imaging Method (fnirs)
(eScholarship, 2023) Demircioglu, Esin Tuna; Girişken, Yener; Çakar, Tuna
Decision-making is a combination of our positive anticipations from the future with the contribution of our past experiences, emotions, and what we perceive at the moment. Therefore, the cues perceived from the environment play an important role in shaping the decisions. Contours, which are the hidden identity of the objects, are among these cues. Aesthetic evaluation, on the other hand, has been shown to have a profound impact on decision-making, both as a subjective experience of beauty and as having an evolutionary background. The aim of this empirical study is to explain the effect of contour types on preference decisions in the prefrontal cortex through risk-taking and aesthetic appraisal. The obtained findings indicated a relation between preference decision, contour type, and PFC subregion. The results of the current study suggest that contour type is an effective cue in decision-making, furthermore, left OFC and right dlPFC respond differently to contour types.
Dog Walker Segmentation
(IEEE, 2022) Ercan, Alperen; Karan, Baris; Çakar, Tuna
In this study dog walkers were separated into clusters according to walkers' walk habits. Due to the fact that the distributions were non-normal, normalization algorithms were applied before the onset of clustering. After normalizing, K Means algorithm and Gaussian Mixture Models used for finding optimum cluster count. According to these clusters, walkers' consecutive months separated to follow-up their behavioral traits. This part of the study adds value to the project to examine walkers' behaviors closer.
Data Deduplication With Adaptive Erasure Code Redundancy (us20160013815a1)
(2016) Arslan, Şuayb Şefik; Wideman, Roderick; Lee, Jaewook; Göker, Turguy
Example apparatus and methods combine erasure coding with data deduplication to simultaneously reduce the overall redundancy in data while increasing the redundancy of unique data. In one embodiment, an efficient representation of a data set is produced by deduplication. The efficient rep resentation reduces duplicate data in the data set. Redundancy is then added back into the data set using erasure coding. The redundancy that is added back in adds protection to the unique data associated with the efficient representation. How much redundancy is added back in and what type of redundancy is added back in may be controlled based on an attribute (e.g., value, reference count, symbol size, number of symbols) of the unique data. Decisions concerning how much and what type of redundancy to add back in may be adapted over time based, for example, on observations of the efficiency of the overall system.
Citation - WoS: 536
Citation - Scopus: 650
Human Semantic Parsing for Person Re-Identification
(2018) Kalayeh, Mahdi M; Başaran, Emrah; Shah, Mubarak; Kamasak, Mustafa E; Gökmen, Muhittin
Person re-identification is a challenging task mainly dueto factors such as background clutter, pose, illuminationand camera point of view variations. These elements hinder the process of extracting robust and discriminative representations, hence preventing different identities from being successfully distinguished. To improve the representation learning, usually local features from human body partsare extracted. However, the common practice for such aprocess has been based on bounding box part detection.In this paper, we propose to adopt human semantic parsing which, due to its pixel-level accuracy and capabilityof modeling arbitrary contours, is naturally a better alternative. Our proposed SPReID integrates human semanticparsing in person re-identification and not only considerably outperforms its counter baseline, but achieves stateof-the-art performance. We also show that, by employinga simple yet effective training strategy, standard populardeep convolutional architectures such as Inception-V3 andResNet-152, with no modification, while operating solelyon full image, can dramatically outperform current stateof-the-art. Our proposed methods improve state-of-the-artperson re-identification on: Market-1501 [48] by ~17% inmAP and ~6% in rank-1, CUHK03 [24] by ~4% in rank-1and DukeMTMC-reID [50] by ~24% in mAP and ~10% inrank-1.
Citation - WoS: 6
Citation - Scopus: 10
Graph-Based Turkish Text Normalization and Its Impact on Noisy Text Processing
(Elsevier, 2022) Topçu, Berkay; Demir, Şeniz
User generated texts on the web are freely-available and lucrative sources of data for language technology researchers. Unfortunately, these texts are often dominated by informal writing styles and the language used in user generated content poses processing difficulties for natural language tools. Experienced performance drops and processing issues can be addressed either by adapting language tools to user generated content or by normalizing noisy texts before being processed. In this article, we propose a Turkish text normalizer that maps non-standard words to their appropriate standard forms using a graph-based methodology and a context-tailoring approach. Our normalizer benefits from both contextual and lexical similarities between normalization pairs as identified by a graph-based subnormalizer and a transformation-based subnormalizer. The performance of our normalizer is demonstrated on a tweet dataset in the most comprehensive intrinsic and extrinsic evaluations reported so far for Turkish. In this article, we present the first graph-based solution to Turkish text normalization with a novel context-tailoring approach, which advances the state-of-the-art results by outperforming other publicly available normalizers. For the first time in the literature, we measure the extent to which the accuracy of a Turkish language processing tool is affected by normalizing noisy texts before being processed. An analysis of these extrinsic evaluations that focus on more than one Turkish NLP task (i.e., part-of-speech tagger and dependency parser) reveals that Turkish language tools are not robust to noisy texts and a normalizer leads to remarkable performance improvements once used as a preprocessing tool in this morphologically-rich language.
Citation - WoS: 1
Citation - Scopus: 1
Identification of Candidate Biomarkers and Pathways in Breast Cancer by Differential Network Analysis
(Inderscience Publishers, 2020) Mendi, Onur; Karahoca, Adem
Breast cancer is one of the most malignant cancers in women worldwide. The aim of the present study was to explore the underlying biological mechanisms of breast cancer. For this purpose, we propose a novel framework to reveal mechanisms that drive disease progression in breast cancer by combining prior knowledge in the literature with differential networking methodology. Our integration framework has resulted in the most important genes and interactions by allowing ranking the breast cancer-specific gene network. YY1, SMARCA5, FOXM1, STAT4 and PTTG1 were found to be the most important genes in breast cancer. Functional and pathway enrichment analyses identified numerous pathways that may play a critical role in disease progression. Considering the success of the comparison of the results with the literature, the systemic lupus erythematosus pathway may be a potential target of breast cancer.
Fault Detection Model Using Measurement Data in Fiber Optic Internet Lines
(IEEE, 2023) Çakar, Tuna; Savaş, Kerem; Battal, Eray; Özkan, Gözde
In this study, a model has been developed to predict potential faults in advance based on performance metrics of various fiber-optic internet lines, as well as alarm (fault data) and performance measurement values from the 5 hours prior to the occurrence of the alarm. Performance metrics that vary over time have been analyzed in a time-series format based on alarm numbers, and anomaly detection methods have been used to label the data for any potential patterns that may occur in the performance metrics specific to the alarm. The labeled data was then fed into a classification model to create a model that enables to detect possible patterns in the relevant performance values for the specific fault type. The best performing model was Random Forest Classifier with accuracy and F1 scores of 0.89 and 0.84 respectively.
Citation - WoS: 6
Citation - Scopus: 7
A Reliability Model for Dependent and Distributed Mds Disk Array Units
(IEEE Transactions on Reliability, 2018) Arslan, Şuayb Şefik
Archiving and systematic backup of large digital data generates a quick demand for multi-petabyte scale storage systems. As drive capacities continue to grow beyond the few terabytes range to address the demands of today’s cloud, the likelihood of having multiple/simultaneous disk failures became a reality. Among the main factors causing catastrophic system failures, correlated disk failures and the network bandwidth are reported to be the two common source of performance degradation. The emerging trend is to use efficient/sophisticated erasure codes (EC) equipped with multiple parities and efficient repairs in order to meet the reliability/bandwidth requirements. It is known that mean time to failure and repair rates reported by the disk manufacturers cannot capture life-cycle patterns of distributed storage systems. In this study, we develop failure models based on generalized Markov chains that can accurately capture correlated performance degradations with multiparity protection schemes based on modern maximum distance separable EC. Furthermore, we use the proposed model in a distributed storage scenario to quantify two example use cases: Primarily, the common sense that adding more parity disks are only meaningful if we have a decent decorrelation between the failure domains of storage systems and the reliability of generic multiple single-dimensional EC protected storage systems.
Citation - Scopus: 2
A Visualization Platfom for Disk Failure Analysis
(IEEE, 2018) Arslan, Şuayb Şefik; Yiğit, İbrahim Onuralp; Zeydan, Engin
It has become a norm rather than an exception to observe multiple disks malfunctioning or whole disk failures in places like big data centers where thousands of drives operate simultaneously. Data that resides on these devices is typically protected by replication or erasure coding for long-term durable storage. However, to be able to optimize data protection methods, real life disk failure trends need to be modeled. Modelling helps us build insights while in the design phase and properly optimize protection methods for a given application. In this study, we developed a visualization platform in light of disk failure data provided by BackBlaze, and extracted useful statistical information such as failure rate and model-based time to failure distributions. Finally, simple modeling is performed for disk failure predictions to alarm and take necessary system-wide precautions.
Predicting Animal Behaviours: Physical and Behavioural Classification Of Dog Walking Levels
(IEEE, 2022) Ozen, Guris; Karan, Baris; Çakar, Tuna
Methods of predicting canine behaviour is an area covered by canine behaviour experts. This study aims to predict the behaviour of dogs during walking based on available information about dogs. In this data-driven project based on up-to-date company data, the problem of predicting dog behaviour was addressed in two different ways. First, it is aimed to create a supervised classification model. Within the scope of this study, improvements were made to various classification algorithms. The results were analyzed in different axes. Secondly, it is aimed to create a new parameter that predicts dog walking difficulties by formulating the parameters.
Citation - WoS: 7
Citation - Scopus: 3
Facial Landmark Localization in Depth Images Using Supervised Ridge Descent
(2015) Camgoz, Necati Cihan; Gökberk, Berk; Akarun, Lale; Struc, Vitomir; Kindiroglu, Ahmet Alp
Supervised Descent Method (SDM) has proven successful in many computer vision applications such as face alignment, tracking and camera calibration. Recent studies which used SDM, achieved state of the-art performance on facial landmark localization in depth images [4]. In this study, we propose to use ridge regression instead of least squares regression for learning the SDM, and to change feature sizes in each iteration, effectively turning the landmark search into a coarse to fine process. We apply the proposed method to facial landmark localization on the Bosphorus 3D Face Database; using frontal depth images with no occlusion. Experimental results confirm that both ridge regression and using adaptive feature sizes improve the localization accuracy considerably.
The Application of Two Bayesian Personalized Ranking Approaches Based on Item Recommendation From Implicit Feedback
(Ieee, 2024) Tagtekin, Burak; Sahin, Zeynep; Çakar, Tuna; Drias, Yassine
The present study has aimed to provide a different ranking approach that will be used actively in a sector-specific application regarding the optimization of item ranking presented to the users. The current online approach in several different applications still holds a manual ranking algorithm whose parameters are determined by the data specialists with adequate domain-knowledge. The obtained findings from the present study indicate that the optimized Bayesian Personalized Ranking models will be used for providing a suitable, data-driven input for the ranking system that would serve to be personalized. The outcomes of the present study also demonstrate that the model using LearnBPR optimized with a stochastic gradient descent algorithm outperform the other similar methods. The sample model outputs were also investigated by a user sample to ensure that the algorithm was working correctly. The next potential step is to provide a normalization process to include the extracted information to the current ranking system and observe the performance of this new algorithm with the A/B tests conducted.
Citation - WoS: 1
Citation - Scopus: 1
Adaptive Boosting of Dnn Ensembles for Brain-Computer Interface Spellers
(IEEE, 2021) Çatak, Yiğit; Aksoy, Can; Özkan, Hüseyin; Güney, Osman Berke; Koç, Emirhan; Arslan, Şuayb Şefik
Steady-state visual evoked potentials (SSVEP) are commonly used in brain computer interface (BCI) applications such as spelling systems, due to their advantages over other paradigms. In this study, we develop a method for SSVEP-based BCI speller systems, using a known deep neural network (DNN), which includes transfer and ensemble learning techniques. We test performance of our method on publicly available benchmark and BETA datasets with leave-one-subject-out procedure. Our method consists of two stages. In the first stage, a global DNN is trained using data from all subjects except one subject that is excluded for testing. In the second stage, the global model is fine-tuned to each subject whose data are used in the training. Combining the responses of trained DNNs with different weights for each test subject, rather than an equal weight, provide better performance as brain signals may differ significantly between individuals. To this end, weights of DNNs are learnt with SAMME algorithm with using data belonging to the test subject. Our method significantly outperforms canonical correlation analysis (CCA) and filter bank canonical correlation analysis (FBCCA) methods.
Eaft: Evolutionary Algorithms for Gcc Flag Tuning
(IEEE, 2022) Tagtekin, Burak; Çakar, Tuna
Due to limited resources, some methods come to the fore in finding and applying the factors that affect the working time of the code. The most common one is choosing the correct GCC flags using heuristic algorithms. For the codes compiled with GCC, the selection of optimization flags directly affects the speed of the processing, however, choosing the right one among hundreds of markers during this process is a resource consuming problem. This article explains how to solve the GCC flag optimization problem with EAFT. Rather than other autotuner tools such as Opentuner, EAFT is an optimized tool for GCC marker selection. Search infrastructure has been developed with particle swarm optimization and genetic algorithm with diffent submodels rather than using only Genetic Algorithm like FOGA. © 2022 IEEE.
Citation - Scopus: 1
Enhancing Quality Control in Plastic Injection Production: Deep Learning-Based Detection and Classification of Defects
(IEEE, 2023) Mutlu, İsmail; Çakar, Tuna; Aslan, Yeşim; Yıldız, Ahmet; Sayar, Alperen; Şimsek, Kamil; Tunalı, Mustafa Mert
This study investigates the applicability of diverse deep learning techniques in detecting and classifying defects within plastic injection manufacturing processes. The findings derived from the models yield several feasible solutions that hold potential practical implications. Notably, the implementation of the Xception model as a classification framework presents a potential domain for enhancing quality control procedures. The developed models, trained on the prepared data sets, provide compelling evidence for the potential utilization of artificial intelligence technologies in the manufacturing industry. Consequently, this study represents a noteworthy contribution to the limited yet auspicious academic research in the field.
Citation - Scopus: 4
Investigation of the Motion of a Spherical Object Located at Soft Elastic and Viscoelastic Material Interface for Identification of Material Properties
(Academic Enhancement Department, King Mongkut's University of Technology North Bangkok, 2024) Körük, Hasan; Pouliopoulos, A.N.
Measuring the properties of soft viscoelastic materials is challenging. Here, the motion of a spherical object located at the soft elastic and viscoelastic material interface for the identification of material properties is thoroughly investigated. Formulations for different loading cases were derived. First, the theoretical models for a spherical object located at an elastic medium interface were derived, ignoring the medium viscosity. After summarizing the model for the force reducing to zero following the initial loading, we developed mathematical models for the force reducing to a lower non-zero value or increasing to a higher non-zero value, following the initial loading. Second, a similar derivation process was followed to evaluate the response of a spherical object located at a viscoelastic medium interface. Third, by performing systematic analyses, the theoretical models obtained via different approaches were compared and evaluated. Fourth, the measured and predicted responses of a spherical object located at a gelatin phantom interface were compared and the viscoelastic material properties were identified. It was seen that the frequency of oscillations of a spherical object located at the sample interface during loading was 10–15% different from that during unloading in the experimental studies here. The results showed that different loading cases have immense practical value and the formulations for different loading cases can provide an accurate determination of material properties in a multitude of biomedical and industrial applications. © 2023 King Mongkut’s University of Technology North Bangkok. All Rights Reserved.
Citation - WoS: 4
Citation - Scopus: 8
Facial Landmark Localization in Depth Images Using Supervised Descent Method
(2015) Gökberk, Berk; Akarun, Lale; Camgoz, Necati Cihan
This paper proposes using the state of the art 2D facial landmark localization method, Supervised Descent Method (SDM), for facial landmark localization in 3D depth images. The proposed method was evaluated on frontal faces with no occlusion from the Bosphorus 3D Face Database. In the experiments, in which 2D features were used to train SDM, the proposed approach achieved state-of-the-art performance for several landmarks over the currently available 3D facial landmark localization methods.

Browse

Recent Submissions