Browsing by Author "Inan, Tolga"

Now showing 1 - 3 of 3

Citation - Scopus: 1
Bit Segmentation of Non-Line of Sight Data in Optical Camera Communication Using U-Net
(Iop Publishing Ltd, 2025) Ozkan, Cagla; Inan, Tolga; Baykal, Yahya
Optical Camera Communication (OCC) utilizes image sensors to decode modulated light signals from light-emitting diodes (LEDs), offering a cost-effective solution for wireless communication. However, data extraction in non-line-of-sight (NLOS) conditions is challenging due to signal distortions caused by obstacles and reflections. Traditional segmentation techniques, such as Otsu's thresholding and adaptive thresholding, are computationally efficient but struggle with lighting variations, background interference, and high-frequency distortions, limiting their effectiveness in real-world OCC applications. To address these limitations, we propose a U-Net convolutional neural network, trained on a diverse dataset covering various camera distances, lighting conditions, and reflection levels to improve segmentation accuracy. The proposed model achieves up to 25% BER improvement, outperforming traditional thresholding methods and ensuring more reliable bit extraction in challenging OCC environments. These advancements make deep learning a promising approach for improving OCC applications such as indoor positioning, smart transportation, and secure optical wireless communication.
Ear Semantic Segmentation in Natural Images With Tversky Loss Function Supported Deeplabv3+ Convolutional Neural Network
(2022) Kacar, Umit; Inan, Tolga
Semantic segmentation is a fundamental problem for computer vision. On the other hand, for studies in the field of biometrics, semantic segmentation is gaining more importance. Many successful biometric recognition systems require a high- performance semantic segmentation algorithm. In this study, we present an effective ear segmentation technique in natural images. A convolutional neural network is trained for pixel-based ear segmentation. DeepLab v3+ network structure, with ResNet-18 as the backbone and Tversky lost function layer as the last layer, has been trained with natural and uncontrolled images. We perform the proposed network training using only the 750 images in the Annotated Web Ears (AWE) training set. The corresponding tests are performed on the AWE Test Set, University of Ljubljana Test Set, and the Collection A of In-The-Wild dataset. For the Annotated Web Ears (AWE) dataset, intersection over union (IoU) is measured as 86.3% for the AWE database. To the best of our knowledge, this is the highest performance achieved among the algorithms tested on the AWE test set.
Citation - WoS: 13
Citation - Scopus: 17
A Machine Learning Study To Enhance Project Cost Forecasting
(Elsevier, 2022) Narbaev, Timur; Hazir, Oncu; Inan, Tolga
In project management it is critical to obtain accurate cost forecasts using effective methods. This study presents a Machine Learning model based on Long-Short Term Memory to forecast the project cost. The model uses the seven-dimensional feature vector, including schedule and cost performance factors and their moving averages as a predictor. Based on the cost variation patterns from the training phase, we validate the model using three hundred experiments in the testing phase. Overall, the proposed model produces more accurate cost estimates when compared to the traditional Earned Value Management index-based model. Copyright (C) 2022 The Authors.