Infrared (IR) imaging sensors, designed to detect the wavelength range between 0.9 mu m and 14 mu m, offer unique advantages over daylight cameras in consumer, industrial, and defense applications. However, IR images lack natural color information and can be challenging for individuals without sensor-specific training to interpret. Consequently, transforming IR images into perceptually realistic color images represents a valuable research endeavor with significant commercial potential. Recently, various studies utilizing deep neural networks for colorizing single-mode (near-IR or thermal) infrared images have been reported. This article will apply a common neural network architecture to images captured with different imaging modes (near-IR, thermal IR, and low-light) for colorization and compare the results. These experiments will examine the influence of perceived wavelength on the colorization process.
This study proposes a speech compression method based on one-dimensional convolutional autoencoder and residual vector quantization. The proposed method offers different compression ratios at low bit rates. Speech quality evaluation metric (PESQ) was used to test the performance of the proposed method. Experimental results show that the proposed method achieves a PESQ value of 1.903 for 2.5 kbps and 2.24 for 5 kbps.
Obstacle avoidance motion planning is a basic problem in mobile robotics. In cases like in a factory, the environment map is known, and the planner’s main task is to find a feedback control law that steers the robot to the goal location without collision. This study presents a novel deterministic method that quickly finds a suitable path by covering the obstacle-free space from start to goal with overlapping polygonal regions guided by Voronoi diagrams. Using a reference governor ensures that the robot stays in these regions and safely moves to the goal. A simulation study confirms the computational gains of our method.
This study aims to compute the periodicities of solar and geomagnetic indices, such as the solar indice IMF-Bz, Dst, Kp, and AE, using the spectral analysis method. This approach offers a fresh perspective on the dynamics of Space Weather and its implications for satellite technology. The primary focus is on investigating the periodic changes in these indices through long-term, one-dimensional temporal Fourier Transform analysis, after the data has been smoothed by applying a sliding median filter. In a solar maximum year, such as 2001, the dominant period for the solar indice IMF-Bz aligns with the Sun’s rotation period, which is approximately 27 days. Additionally, the upper and lower harmonics of this cycle have been identified in indices that reflect levels of geomagnetic disturbance, occurring at intervals of approximately 1 month, 3 months, and 6 months, as well as shorter periods of 1 week, 2 weeks, and 3 weeks, respectively.
This study introduces a method for breast cancer detection using a microstrip patch antenna designed to exploit dielectric property contrasts between healthy and malignant tissues. Three-layered breast phantoms, mimicking human breast tissue, were developed and tested using the antenna for cancer detection. Utilizing a 28.8 x 28.8 mm(2) patch antenna on an FR4 structure, the study examines tumor impact through return loss values. Initial simulations on a multilayer breast phantom model, created with CST Microwave Studio, were conducted for tumor identification. Subsequently, realistic breast phantoms comprising skin, fat, and glandular tissues were fabricated to validate the antenna's performance. Return loss measurements on fabricated phantoms, conducted with a NanoVNA-F Network Analyzer, closely matched simulated results. The study proposes a promising method for breast cancer detection through valuable insights provided to researchers via the testing of the proposed antenna and realistic breast phantoms.
This study explores the potential applications across various domains, including earthquake detection and warning systems, where the system’s sensitivity to ground vibrations can contribute to early seismic event detection. Additionally, the study paves the way of developing applications of VLC/T in mechanical vibration and stability analysis of engines and platforms, offering insights into structural integrity and performance optimization. These multifaceted applications underscore the adaptability and potential of VLC/T systems in diverse fields, heralding advancements in sensing, communication, and security technologies. To achieve this, in this study, Peak to Average Power Ratio (PAPR) is proposed to represent the impact of mechanical shocks and vibrations generated by several weights dropped onto the platform with which the receiver is fixed. Even though non-contact measurement methodology is preferred for various reasons, the proposed measurement campaign obtains the data in contact form; however, the system and signal model proposed in this study could easily be extended into non-contact form. Considering the fact that the proposed measurement campaign employs off-the-shelf products, it is cost-effective and very scalable. 1 1 This manuscript has been authored in part by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doepublic-access-plan).
The natural stone industry holds a billion-dollar economy worldwide and is recognized as a significant component of economic development. Marble blocks, a crucial part of this industry, find widespread usage in fields such as architecture, construction, interior decoration, and sculpture. The quality and characteristics of marble blocks are vital for end-users and commercial suppliers. Analyzing these features plays a critical role in industrial processes. Traditional methods of marble block analysis are time-consuming, costly, and sometimes yield subjective results. Therefore, the utilization of technologies like artificial intelligence and machine learning offers a new perspective in industrial applications. The Look Marble project is a comprehensive research and development initiative designed to execute complex quality assessment processes within the marble industry. The project aims to optimize quality classification, pricing, and marketing strategies throughout the entire lifecycle of marble blocks, from production to sales. Evaluating the heterogeneous characteristics of marble blocks for quality classification and formulating more effective marketing strategies are among the main objectives of the project. This study focuses on the automatic detection of cracks on marble surfaces, a significant module of the Look Marble project. Crack detection is a crucial factor influencing the quality of marble blocks, and the development of this technology is deemed capable of enhancing efficiency and improving quality control processes in the industry.
Robot assisted surgery is quite useful method for medical operations. To use this method properly, evaluation of surgeon with low experience has critical significance. Currently, evaluation is made by simulation programs that specified for robot assisted surgery or result of scoring with a survey made by expert surgeon who tracks all the session. Aim of this study is assessment of surgeons in real life robot-assisted operations with eye-tracker device without an evaluator. To make this evaluation, output with constructed regression models with derived features from an open source data that has been collecting during robot-assisted-surgery simulation is compared with score produced by simulation. Mean of root mean squared error of Xgboost that has 0.124 value and R2 score as 0.456 value were obtained for best model output.
Forests are terrestrial ecosystems that provide ecological balance and environmental benefits. Due to uncontrolled logging, increasing population, and commercial use, forest reserves are rapidly decreasing worldwide. This situation carries risks that will lead to climate crisis and biodiversity loss. For this reason, trees need to be protected and kept under control. Since human-powered control efforts are insufficient, the use of satellite images and artificial intelligence-based systems has come to the fore to prevent forest destruction. Nowadays, there are studies on tree detection applications from satellite images using different deep learning models. This article aims to combine Swin transformers and YOLO models with a post-processing technique known as weighted boxes fusion to enhance object detection capabilities. The object detection approaches of the two models are different. When the wide variation detection ability of the Swin transformers method is combined with the low variation but sharp detection ability of the YOLO model, the results combine the advantages of both models in tree detection. In this way, the proposed method detected trees with %3.7 higher precision than Swin transformers and %13.1 higher precision than YOLO.
To address the resource gap in open source Turkish extractive summarization datasets, we present XTINGE-SUMExt, MLSUM-TR-Ext, and TES datasets. These datasets were enriched with measurable analyses. We also report metrics regarding these published datasets. These datasets are designed to facilitate the development and evaluation of different summarization strategies.
In sign languages, where communication is achieved through hand gestures, facial expressions and body language, signs are the subject of many studies due to the diversity in terms of the position of different body parts. These diversities are also encountered in emotion detection in Turkish Sign Language (TID), making direct translation of hand gestures inadequate for emotion detection. Accordingly, in this study, for the first time in the literature, sentiment analysis in TID was performed using facial expressions and hand gestures. For this purpose, a specialized model for the tasks of emotion extraction from facial expressions and gesture recognition from hand gestures was fine-tuned with the dataset collected in this study. As a result, facial expressions are found to be more significant than hand gestures in sentiment analysis in TID, but when supported with hand gestures, the performance improved even more.
The digitalization process is progressing at a very high speed all over the world. While this situation provides many conveniences in today’s life, it also brings along a problem such as analyzing and processing the huge digital data. This also applies to published academic studies. In this sense, the process of evaluating each study to access previously unknown information within the studies requires a very laborious process. For this reason, in this study, the publications obtained for the target diseases were analyzed by text analysis processes and converted into a graph structure that enables the linking of meaningful terms through biomedical relationships. On the dense graph structure obtained, binary biomedical entities with important links such as treats, causes, associated_with were queried. The entity pairs obtained according to the query results were also confirmed by manual search method and proved to be real connections. In this study, retrieval of known biomedical entities with the proposed approach solved the time-consuming manual search problem. There is also the potential to obtain unknown/unexplored possible new relationships (e.g., therapeutic, causal, etc.) with multiple binary linking patterns.
Graf’s ultrasonography (US) method is one of the most commonly used imaging techniques for developmental dysplasia of the hip (DDH) and is universally accepted for the assessment of neonatal hips [1]. However, the training process is lengthy and requires supervision until the evaluator achieves expertise. Computer-based segmentation and object detection tools may assist less experienced evaluators in identifying anatomical structures and classifying hip US images. This method involves anatomical description as well as measuring bone and soft tissue coverage in coronal two-dimensional (2D) US images of the hip. During scanning with the ultrasound probe, the physician has to decide whether the image is in the standard plane and whether the image is measurable. An image must contain a straight iliac wing, lower limb of the ilium, and the labrum to be classified as measurable [2, 3]. Graf’s method is prone to interpreter variability due to the anatomical complexity of the hip structures, which can lead to misclassification [4]. When anatomical regions are not precisely identified, the selection of points for angle calculations may not be accurately determined, rendering the image unacceptable for measurements. This study comparatively measured success using a different model of the YOLOv8 algorithm to detect the labrum, lower limb of the ilium, and iliac wing regions in 200 measurable hip ultrasonography images obtained in the standard plane. With the YOLOv8x configuration, the labrum, ilium, and acetabulum were detected with success rates of 93.57%, 98.30% and 94.25% respectively, with an intersection over union (IoU) of 0.25. Our findings indicate that the YOLOv8x-based algorithm shows significant promise for the detection of labrum, ilium, and iliac wing regions in the standard plane.
In this study, a new topology that realizes the behavior of a meminductor without the use of external passive elements is proposed. This topology consists of two integrators and an operational transconductance amplifier. When compared with similar circuits presented in the literature, it is predicted that the proposed circuit occupies less area on the entire chip than circuits implemented with external passive elements. To demonstrate the feasibility of the circuit, simulation results are obtained using 0.18 mu m CMOS parameters in the Cadence design environment. These findings confirm the accuracy of the theoretical analysis and the effective applicability of the proposed structure
Sustainability plays a central role in addressing global challenges, encompassing environmental, economic, and social principles. Sustainability reports evaluate organizations’ environmental impacts, social responsibilities, and adherence to governance principles, ensuring transparency. In this study, Retrieval Augmented Generation (RAG) technology is employed to assess the Turkish sustainability reports of 10 companies listed in the BIST Sustainability-25 index in terms of ESG (Environmental, Social, Governance) factors. Information retrieval is conducted producing 47 prompts related to ESG criteria. The BM25 and BERTurk approaches identify the most relevant sections in sustainability reports, which are then input into the GPT3.5- Turbo model. According to the results obtained, it is seen that using the BM25 method performed better than BERTurk when extracting information from documents with the RAG approach
With the advancing sensor (receiver-processor) technology, Time Difference of Arrival (TDOA) based systems have become attractive for estimating the direction of a transient or pulsed signal. Compared to traditional phase or amplitude comparison-based direction finding (DF) systems, distributed TDOA systems have several advantages. These include the absence of uncertainty problems, direction of arrival (DOA) estimation independent of sensor and signal characteristics, mobility, and the dynamic array manifold structure. On the other hand, in distributed TDOA systems, a synchronization problem arises among clocks since each node has its own local clock. This article introduces a deep neural network (DNN) structure capable of estimating the direction of a signal in the presence of time offset. Experimental (simulation) results demonstrate that the proposed method outperforms classical approaches.
Autism is a neurodevelopmental disorder that typically begins in childhood and continues throughout life. Autism Spectrum Disorder (ASD) exhibits prominent features such as difficulties in social interaction, communication problems, repetitive behaviors, and restricted interests. Diagnosis is usually made using methods like clinical observation, developmental screening tools, psychological assessments, and behavioral tests, although methods like Electroencephalography (EEG) may also be used in rare cases. Artificial intelligence (AI) is increasingly playing a significant role in the diagnosis and evaluation of ASD. In the literature, there are many advanced methods that have the potential to be used for ASD diagnosis using EEG and artificial intelligence. In this study, a novel approach for EEG-based ASD detection was developed using a multi-input dimensional Convolutional Neural Network (CNN) model. The developed method was tested on an openly accessible dataset obtained from King Abdulaziz University Hospital, achieving an accuracy of 92.93%.
In recent years, security cameras have been used intensively both in our country and in the world for reasons such as obtaining intelligence or obtaining an event recording. These cameras can be fixed as well as PTZ (pan-tilt-zoom)- capable. PTZ cameras are capable of tilting, panning and zooming with remote management, but the movement and optical capabilities of each brand and model of PTZ camera may differ. In this study, a new method has been developed to automatically calibrate cameras onsite precisely with different mobility and then direct them to the desired location for a specific purpose.
People who undergo upper extremity amputation for various reasons face a dramatic decrease in their quality of life. Prostheses designed to remedy this situation, at least partially, are becoming more successful day by day. In this study, the “Nvidia Jetson“ card-supported experiment platform, designed especially for the development of artificial intelligence procedures in under-actuated myoelectric hand prostheses, is presented. The system, in which electromyography (EMG) signals are processed with artificial intelligence procedures and converted into control signals, is designed modularly so that different methods and equipment can be tested. The platform, where all the subunits that form the whole system are combined and their interactions are ensured, was tested with a support vector machine (SVM) based motion classification procedure selected as an example and the results were examined. In the design output, it was seen that hand control was achieved with 92.5% accuracy in five hand movements. Using this platform, different algorithms will be tested, and performance will be increased with the hardware and training used
Numerous deep learning-based methods have been proposed for achieving high accuracy in 3D object reconstruction. However, when examining the recent models, we observe that their performances are very close. We claim that more detailed evaluation methods are needed to broaden the comparisons and allow new research directions. Accordingly, in this study, we propose a novel benchmark to evaluate at the part level over three state-of-the-art reconstruction models using the novel rich dataset, 3DCoMPaT++. To evaluate holistic shape reconstruction outputs at the part level, the Part F-Score metric is proposed. Adapting a dataset proposed from a close domain is important for enabling new data to 3D object reconstruction applications and for guiding new adaptations.
