Scopus İndeksli Yayınlar Koleksiyonu
Permanent URI for this collectionhttps://hdl.handle.net/20.500.12573/395
Browse
3 results
Search Results
Article Frequency-Based Deep Occlusion Awareness Instance Segmentation(MDPI, 2026-02-26) Guzel, Yasin; Aydin, Zafer; Talu, Muhammed FatihOne major challenge faced by deep learning-based methods that detect target objects in the form of bounding boxes is object occlusion. High degrees of occlusion significantly diminish the accuracy of instance segmentation. Nonetheless, complex-valued Fourier descriptors can robustly represent object boundaries using minimal information. In this study, the impact of integrating Fourier descriptors-renowned for their strong representational capacity-with deep network models (UNet) that exhibit high generalization performance on instance segmentation accuracy was investigated. Within the scope of the research, nine network models were designed based on different strategies for utilizing frequency components. These variants fall into four strategy families: (i) UNet-style spectrum regression on fixed low-frequency windows (FUNet), (ii) magnitude-guided frequency selection/ROI construction (FUNet-Thr, FUNet-BBox), (iii) sequence models over tokenized FFT coefficients (BiLSTM Patch/Sorted), and (iv) encoder-only spectrum predictors with different depth/capacity (EncoderFFT1/2). To fairly evaluate the models' performance in segmenting objects subjected to disruptive factors (e.g., occlusion, blurring, noise), a specialized synthetic dataset was prepared. The task is formulated as single-target (single-instance), single-class segmentation. This dataset, automatically generated according to initial parameter values, contains images of objects moving at various speeds within a single frame. Among these models, the one termed FUNet, which relies on partial matching of central frequency components, achieved the highest segmentation accuracy despite the disruptive effects. Under the challenging Dataset 8 setting, the proposed FUNet achieved the highest overlap-based performance (Dice = 0.9329, IoU = 0.8842) among Attention U-Net, U-Net, and FourierNet, with statistically significant gains confirmed by paired per-image tests.Article Citation - WoS: 9Citation - Scopus: 12Not All Emerging Markets Are the Same: A Classification Approach With Correlation Based Networks(Elsevier Science inc, 2017-12) Sensoy, Ahmet; Ozturk, Keyser; Hacihasanoglu, Erk; Tabak, Benjamin M.Using dynamic conditional correlations and network theory, this study brings a novel interdisciplinary framework to define the integration and segmentation of emerging countries. The individual EMBI+ spreads of 13 emerging countries from January 2003 to December 2013 are used to compare their interaction structure before (phase 1) and after (phase 2) the global financial crisis. Accordingly, the unweighted average of dynamic conditional correlations between cross country bond returns significantly increases in phase 2. At first glance, the increased co-movement degree suggests an integration of the sample countries after the crisis. However, using correlation based stable networks, we show that this is not enough to make such a strong conclusion. In particular, we reveal that the increased average correlation is more likely to be caused by clusters of countries that exhibit high within-cluster co-movement but not between-cluster co-movement. Potential reasons for the post-crisis segmentation and important implications for international investors and policymakers are discussed. (C) 2016 Elsevier B.V. All rights reserved.Conference Object Citation - Scopus: 1Comparison of Lung Tumor Segmentation Methods on PET Images(IEEE, 2015-10) Eset, Kubra; Icer, Semra; Karacavus, Seyhan; Yilmaz, Bulent; Kayaalti, Omer; Ayyildiz, Oguzhan; Kaya, EserLung cancer is the most common cause of cancer-related deaths that occur all over the world. Recently, various image processing approaches have been used on PET images in order to characterize the uniformity, density, coarseness, roughness, and regularity (i.e., texture properties) of the intratumoral F-18-fluorodeoxyglucose (FDG) uptake. The first and important step of this kind of analysis is to differentiate tumor region from other structures and background, which is called segmentation. In this study, k-means, active contour (snake), and Otsu's tresholding methods were applied on PET images obtained from 36 patients and the performances were compared by the nuclear medicine expert in our team. The results show that Otsu tresholding approach is more selective.