연구성과로 돌아가기
2024 연구성과 (169 / 286)
※ 컨트롤 + 클릭으로 열별 다중 정렬 가능합니다.
Excel 다운로드
| WoS | SCOPUS | Document Type | Document Title | Abstract | Authors | Affiliation | ResearcherID (WoS) | AuthorsID (SCOPUS) | Author Email(s) | Journal Name | JCR Abbreviation | ISSN | eISSN | Volume | Issue | WoS Edition | WoS Category | JCR Year | IF | JCR (%) | FWCI | FWCI Update Date | WoS Citation | SCOPUS Citation | Keywords (WoS) | KeywordsPlus (WoS) | Keywords (SCOPUS) | KeywordsPlus (SCOPUS) | Language | Publication Stage | Publication Year | Publication Date | DOI | JCR Link | DOI Link | WOS Link | SCOPUS Link |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ○ | ○ | Article | Gem5-AVX: Extension of the Gem5 Simulator to Support AVX Instruction Sets | Recent commodity x86 CPUs still dominate the majority of supercomputers and most of them implement vector architectures to support single instruction multiple data (SIMD). Although research on architectural exploration requires computer architecture simulators and a number of simulators have been developed, only a few tools support recent x86 SIMD instructions. This paper describes gem5-AVX, an extended version of the gem5 simulator that enables simulating recent x86 SIMD extensions, especially targeted for high performance computing (HPC). The gem5-AVX comprises advanced vector extension (AVX), AVX2 and subsets of AVX-512, except for cache and memory management instructions. Moreover, it covers full set of streaming SIMD extensions (SSE) and subsequent extensions that are required to simulate HPC workloads. It can simulate the key features of the AVX, AVX2 and AVX-512 such as 256 and 512 bits wide registers, three and four operands syntax, fused multiply-add (FMA), vector gather-scatter using vector scale-index-base (VSIB), mask registers, embedded broadcasting, compressed displacement memory addressing mode. We evaluate the accuracy of gem5-AVX by comparing its results to those of real hardware and Intel's software development emulator (SDE) running benchmark suites,i.e., high-performance linpack (HPL), high-performance conjugate gradient (HPCG) and NAS parallel benchmark (NPB) which are representative programs in the HPC field. The gem5 and gem5-AVX are compared with the speed-up of HPL benchmark according to configuration combinations. Gem5-AVX, with mean absolute percentage errors of 7.3-9.2% and 9.2-11.9%, is more accurate than gem5, which shows mean absolute percentage errors 17.9-21.5% and 19.7-29.7% for Haswell and Skylake processors, respectively. | Lee, Seungmin; Kim, Youngsok; Nam, Dukyun; Kim, Jong | Pohang Univ Sci & Technol POSTECH, Dept Comp Sci & Engn, Pohang 80305, South Korea; Yonsei Univ, Dept Comp Sci, Seoul 03722, South Korea; Kyungpook Natl Univ, Sch Comp Sci & Engn, Daegu 41566, South Korea; Korea Inst Sci & Technol Informat KISTI, Daejeon 34141, South Korea | ; Nam, Dukyun/AFX-2852-2022 | 57196320582; 56259399400; 35805721000; 59643949800 | dynam@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0 | 2025-04-16 | 0 | 2 | Gem5 simulator; x86 SIMD; AVX; AVX2; AVX-512 | AVX; AVX-512; AVX2; Gem5 simulator; x86 SIMD | Benchmarking; Cache memory; Memory architecture; Program processors; Software design; Supercomputers; Support vector machines; Syntactics; Advanced vector extension; Advanced vector extension-512; AVX2; Benchmark testing; Code; Decoding; Gem5 simulator; Instruction set; Memory-management; Multiple data; Support vectors machine; X86 single instruction multiple data; Vectors | English | 2024 | 2024 | 10.1109/access.2024.3359296 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | Genetic Algorithm Guided Image Channel Selection for Skin Lesion Segmentation | Skin cancer stands as the most widespread form of cancer globally, and early detection significantly enhances its treatability. Even though deep learning techniques have greatly improved the precision of segmentation, there is still potential for enhancement by tackling substantial challenges like the variability in lesion sizes, colors, shapes, and differences in contrast levels. This paper introduces an innovative approach to Genetic Algorithm-guided feature selection in skin lesion segmentation. The Modified TMU-Net (Transformer Meets U-Net) architecture overcomes limitations by replacing the ResNet block with a custom block for improved adaptability to variable input sizes. The dual-pipeline design integrates transformers for global representations and spatial dependency with convolutional neural networks for local contextual representations. A Genetic Algorithm (GA) is employed alongside the architecture to optimize image channel selection for enhanced segmentation. The GA iteratively refines solutions encoded as binary vectors, representing combinations of image types. This framework combines the original RGB images with data derived from the principles of skin illumination and imaging. Our approach involves incorporating data from distinct color bands, grayscale images immune to variations in illumination, and images with reduced shading effects. The combination of R-n, GRAY, SA, and RGB features produces qualitatively superior results compared to the RGB images. The proposed model was trained on the ISIC 2017 publicly available dataset and achieved successful outcomes. | Jasurbek, Nazarov; Ivan, Dzeuban Fenyom Fenyom; Ajani, Oladayo S.; Mallipeddi, Rammohan | Kyungpook Natl Univ, Dept Artificial Intelligence, Daegu 41566, South Korea | AJANI, Oladayo/HIR-9607-2022; Mallipeddi, Rammohan/AAL-5306-2020 | 59337493200; 59372052400; 57465126000; 25639919900 | mallipeddi.ram@gmail.com; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0.77 | 2025-05-07 | 3 | 4 | Skin; Image segmentation; Lesions; Image color analysis; Lighting; Feature extraction; Gray-scale; Semantic segmentation; Genetic algorithms; Dermatology; Transformer; semantic segmentation; genetic algorithm; dermoscopic images; attention | CANCER STATISTICS; CLASSIFICATION; DIAGNOSIS | attention; dermoscopic images; genetic algorithm; semantic segmentation; Transformer | Binary images; Convolutional neural networks; Deep learning; Feature Selection; Image enhancement; RGB color model; Attention; Channel selection; Dermoscopic images; Genetic algorithm algorithm; Image channels; Lesion segmentations; RGB images; Semantic segmentation; Skin lesion; Transformer; Semantic Segmentation | English | 2024 | 2024 | 10.1109/access.2024.3462939 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||
| ○ | ○ | Article | Implementation of Grover’s Iterator for Quantum Searching With an Arbitrary Number of Qubits | Grover's algorithm harnesses the power of quantum computing to swiftly locate specific elements in an unstructured database, outperforming classical computers in tasks like database searching. This algorithm capitalizes on the unique ability of qubits to be in both 0 and 1 states simultaneously (superposition), allowing it to scan the entire search space at once. It then boosts the probability of the target element, making it more prominent. While the foundational concepts of Grover's algorithm are well-documented, practical implementation using quantum operators, especially for large search spaces, remains less explored beyond basic examples with a small number of qubits. Existing general synthesis techniques often involve numerous operators or are time-consuming. Our proposed methods specifically address the amplitude-amplification component of Grover's algorithm for any size of search space. These methods detail the required types and quantities of qubits and operators, emphasizing minimal usage and efficient assembly. Developed and evaluated in Python, our methods consistently identified target elements with over 95% accuracy and achieved configurations comparably compact as those in the existing literature but at a faster pace. We anticipate that these methods facilitate practical implementations of Grover's algorithm across various domains. | Lee, Sihyung; Nam, Seung Yeob | Kyungpook Natl Univ, Sch Comp Sci & Engn, Daegu 41566, South Korea; Yeungnam Univ, Dept Informat & Commun Engn, Gyongsan 38541, South Korea | ; Nam, Seung/Q-7486-2019 | 15623380100; 7402276352 | sihyunglee@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0.86 | 2025-04-16 | 1 | 4 | Qubit; Quantum computing; Vectors; Programming; Databases; Synthesizers; Search problems; Grover's algorithm; program synthesis; quantum algorithm; quantum computing; quantum program | Grover's algorithm; program synthesis; quantum algorithm; quantum computing; quantum program | Quantum computers; Quantum optics; Search engines; Grov’s algorithm; Program synthesis; Programming; Quantum algorithms; Quantum Computing; Quantum program; S-algorithms; Search problem; Synthesizer; Database systems | English | 2024 | 2024 | 10.1109/access.2024.3380198 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | Improvement of the Symmetry and Linearity of Synaptic Weight Update by Combining the InGaZnO Synaptic Transistor and Memristor | Obtaining symmetrical and highly linear synapse weight update characteristics of analog resistive switching devices is critical for attaining high performance and energy efficiency of the neural network system. In this work, based on the two-terminal one transistor-one memristor (1T1M) block, the improvement of the symmetry and linearity of synaptic weight update is demonstrated by combining the InGaZnO synaptic transistor and memristor. Due to the symmetric and linear weight update characteristic, a pattern recognition accuracy of 88% is achieved after 50 epochs in the on-chip learning simulation of the hand-written digit images (MNIST) data set. The proposed 1T1M device saves the hardware burden and additional power consumption required to implement non-identical programming pulses. | Yang, Tae Jun; Cho, Jung Rae; Lee, Hyunkyu; Lee, Hee Jun; Myoung, Seung Joo; Lee, Da Yeon; Choi, Sung-Jin; Bae, Jong-Ho; Kim, Dong Myong; Kim, Changwook; Woo, Jiyong; Kim, Dae Hwan | Kookmin Univ, Sch Elect Engn, Seoul 02707, South Korea; Kyungpook Natl Univ, Sch Elect Engn, Daegu 41566, South Korea | KIM, HYUNGTAK/P-7054-2015; Lee, Heejun/GYU-5361-2022; Bae, Jong-Ho/V-5237-2019; Choi, Sung-Jin/ACC-8335-2022; Kim, Dong/W-5846-2019 | 57428906500; 58028344800; 57075153200; 57836279300; 58017933900; 58719793400; 7408120164; 55339347700; 34975080400; 57202353919; 53985749100; 57198637496 | ncwkim@kookmin.ac.kr;jiyong.woo@knu.ac.kr;drlife@kookmin.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 1.29 | 2025-04-16 | 6 | 7 | InGaZnO thin-film transistors; nalog resistive switching synapse; symmetric and linear synaptic weight update; synaptic transistor; memristor; neural network | TEMPERATURE; MEMORY | analog resistive switching synapse; InGaZnO thin-film transistors; memristor; neural network; symmetric and linear synaptic weight update; synaptic transistor | Energy efficiency; Gallium compounds; Memristors; Neural networks; Pattern recognition; Thin film circuits; Zinc compounds; Analog resistive switching synapse; C. thin film transistor (TFT); InGaZnO thin-film transistor; Memristor; Neural-networks; Resistive switching; Symmetric and linear synaptic weight update; Symmetrics; Synaptic transistor; Synaptic weight; Weight update; Thin film transistors | English | 2024 | 2024 | 10.1109/access.2024.3366224 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||
| ○ | ○ | Article | Low-Power Lane Detection Unit With Sliding-Based Parallel Segment Detection Accelerator for FPGA | Recently, with the development of semiconductors and VLSI (Very Large Scale Integrated Circuit), the technology required for autonomous driving is rapidly developing. One of the technologies that cannot be left out is the lane detection function. Lane recognition requires a lot of data from the camera sensor. As a result, the data size increases, making it difficult to process on a lightweight embedded board. This paper proposes a sliding-based parallel segment image processing method to solve this problem. Most boards in autonomous vehicles are lightweight, so the technique has been designed to reduce computation and power consumption. After fetching the image's pixel data, grayscale conversion, Gaussian smoothing, Sobel operator, non-maximum suppression, and hysteresis are performed in parallel. Lanes were detected by performing a Hough transform operation on an image for which edge detection was completed in parallel. Due to the nature of parallel processing, it is more effective when image input is continuous and numerous than single image processing. This algorithm is written in C language and VHDL (VHSIC Hardware Description Language) for two parts in the board, DE1-SoC, FPGA (Field Programmable Gate Array) and HPS (Hard Processor System. Due to the use of the C language and VHDL, parallel programming uses 3.1 times less time, twice as much memory and slightly more power than sequential programming. For hardware languages such as Verilog, the computation algorithms have been converted to a fixed point. When comparing HPS and FPGA, the FPGA consumed significantly fewer resources, with 18 times shorter run time, 50 times fewer clock cycles, 3 times less power, and 183 times less energy. This provides a substantial benefit. | Yun, Heuijee; Park, Daejin | Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu 41566, South Korea | Yun, Heuijee (Heejee)/GOJ-9000-2022 | 57222516795; 55463943600 | boltanut@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 1.72 | 2025-04-16 | 3 | 6 | Autonomous driving; lane detection; canny edge detection; hough transform; FPGA acceleration; low power design | Autonomous driving; canny edge detection; FPGA acceleration; hough transform; lane detection; low power design | Application specific integrated circuits; C (programming language); Computer hardware description languages; Deep learning; Edge detection; Electric power supplies to apparatus; Feature extraction; Field programmable gate arrays (FPGA); Image segmentation; Logic gates; Low power electronics; Neural networks; Pipeline processing systems; System-on-chip; Autonomous driving; Canny edge detection; Convolutional neural network; Deep learning; Field programmable gate array; Field programmable gate array acceleration; Field programmables; Hardware; Image edge detection; Lane detection; Low-power design; Programmable gate array; Hough transforms | English | 2024 | 2024 | 10.1109/access.2023.3348478 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | Low-Power Scalable TSPI: A Modular Off-Chip Network for Edge AI Accelerators | In this paper, we present a novel off-chip network architecture, the Tile Serial Peripheral Interface (TSPI), designed for low-power, scalable edge AI accelerators. Our approach modifies the conventional SPI to support a modular network structure that facilitates the scalable connection of multiple accelerators. The TSPI network employs a subset mapping algorithm for efficient routing and integrates the message passing interface (MPI) protocol to ensure rapid data distribution and aggregation. This modular architecture significantly reduces power consumption and improves processing speed. Experimental results demonstrate that our proposed TSPI network achieves a 54.7% reduction in power consumption and an 82.3% decrease in switching power compared to traditional SPI networks, along with a 23% increase in processing speed when utilizing 16 nodes. These advancements make the TSPI network an effective solution for enhancing AI performance in edge computing environments. | Park, Seunghyun; Park, Daejin | Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu 41566, South Korea; Kyungpook Natl Univ, Sch Elect Engn, Daegu 41566, South Korea | 57903951400; 55463943600 | boltanut@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0 | 2025-05-07 | 0 | 0 | Random access memory; Computational modeling; Performance evaluation; Edge AI; Distributed databases; Scalability; Indexes; Low power electronics; Network architecture; Accelerator architectures; TSPI; off-chip network; edge device; subset mapping algorithm; low power; MPI | edge device; low power; MPI; off-chip network; subset mapping algorithm; TSPI | Edge computing; Gluing; Mapping; System-on-chip; Edge device; Low Power; Mapping algorithms; Message passing interface; Message-passing; Off-chip; Off-chip network; Serial peripheral interface; Subset mapping algorithm; Tile serial peripheral interface; Message passing | English | 2024 | 2024 | 10.1109/access.2024.3466965 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||||
| ○ | ○ | Article | Multi-Task Learning With Self-Defined Tasks for Adversarial Robustness of Deep Networks | Despite the considerable progress made in the development of deep neural networks (DNNs), their vulnerability to adversarial attacks remains a major hindrance to their practical application. Consequently, there has been a surge of interest and investment in researching adversarial attacks and defense mechanisms, with a considerable focus on comprehending the properties of adversarial robustness. Among these intriguing studies, a couple of works show that multi-task learning can enhance the adversarial robustness of DNNs. Based on the previous works, we propose an efficient way to improve the adversarial robustness of a given main task in a more practical multi-task learning scenario by leveraging self-defined auxiliary task. The core concept of our proposed approach lies not just in jointly training predefined auxiliary tasks but in manually defining auxiliary tasks based on the built-in labels of given data, which enables users to efficiently perform multi-task learning without the need for pre-defined auxiliary tasks. The newly generated self-defined tasks remain "hidden" from attackers and serve a supplementary role in improving the adversarial accuracy of the main task. In addition, the hidden auxiliary tasks also enable to build a rejection module that utilizes predictions from the auxiliary tasks to enhance the reliability of the prediction results. Through experiments conducted on five benchmark datasets, we confirmed that multi-task learning with self-defined hidden tasks can be actively employed to enhance the adversarial robustness and reliability. | Hyun, Changhun; Park, Hyeyoung | Kyungpook Natl Univ, Sch Comp Sci & Engn, Daegu 41566, South Korea | ; HYUN, CHANGHUN/MYR-4076-2025 | 57188754151; 55713613500 | hypark@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0 | 2025-04-16 | 1 | 1 | Task analysis; Multitasking; Robustness; Training; Predictive models; Computational modeling; Perturbation methods; Adversarial machine learning; Artificial neural networks; Adversarial attack; adversarial robustness; multi-task learning; adversarial training; self-defined auxiliary tasks | Adversarial attack; adversarial robustness; adversarial training; multi-task learning; self-defined auxiliary tasks | Job analysis; Perturbation techniques; Adversarial attack; Adversarial robustness; Adversarial training; Computational modelling; Multitask learning; Perturbation method; Predictive models; Robustness; Self-defined auxiliary task; Task analysis; Deep neural networks | English | 2024 | 2024 | 10.1109/access.2024.3355924 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | Numerical Analysis of Ion Flow in One-Bipole HVDC Transmission Line Using Revised Charge Injection Methods | In the evolving field of electric power transmission networks, high-voltage direct current (HVDC) transmission has garnered attention for its efficacy in long-distance power delivery. However, HVDC systems are susceptible to corona discharges, which generate ions that disrupt the electric field distribution and pose safety concerns. To address these challenges, this study introduces new calculation techniques for predicting the electric field and ion current density around HVDC transmission lines using the finite-element method. The onset fields for the corona discharge were established at 14 and 13 kV/cm for positive and negative ions, respectively. Three novel techniques-average (A), cosine (C), and average-cosine combination (AC)-were introduced for continuous charge distribution. Additionally, an enhancement factor beta was incorporated to reflect the various climatic conditions, enhancing the model's adaptability. This approach streamlines the analysis by reducing the reliance on complex parameters such as conductor roughness coefficient and climate constants. The techniques were validated across four different bundle configurations of transmission lines, with the AC technique demonstrating superior accuracy in predicting the electric field and ion current density, affirming its robustness in diverse scenarios, including under wind conditions. This research marks a significant advancement in modeling electrical discharge phenomena in HVDC environments, providing a simplified yet precise tool for ensuring electrical safety. | Kim, Minhee; Lee, Se-Hee | Korea Electrotechnol Res Inst, Ecofriendly Power Apparat Res Ctr, Chang Won 51543, South Korea; Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu 41566, South Korea | Kim, Minhee/LRT-2058-2024 | 57214228950; 55009905600 | shlees@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0.86 | 2025-04-16 | 2 | 3 | Corona discharge; HVDC; finite-element method; corona onset; charge injection | FINITE-ELEMENT-ANALYSIS; FIELD; DENSITY | charge injection; Corona discharge; corona onset; finite-element method; HVDC | Current density; Electric corona; Electric power transmission networks; Finite element method; HVDC power transmission; Negative ions; Numerical methods; Conductor; Corona; Corona discharges; Corona onset; Direct current transmissions; Discharge (electric); Field currents; Finite element analyse; High-Voltage Direct Current; Transmission-line; Electric lines | English | 2024 | 2024 | 10.1109/access.2024.3361937 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||
| ○ | ○ | Article | Optimizing Prompts Using In-Context Few-Shot Learning for Text-to-Image Generative Models | Recently, various text-to-image generative models have been released, demonstrating their ability to generate high-quality synthesized images from text prompts. Despite these advancements, determining the appropriate text prompts to obtain desired images remains challenging. The quality of the synthesized images heavily depends on the user input, making it difficult to achieve consistent and satisfactory results. This limitation has sparked the need for an effective prompt optimization method to generate optimized text prompts automatically for text-to-image generative models. Thus, this study proposes a prompt optimization method that uses in-context few-shot learning in a pretrained language model. The proposed approach aims to generate optimized text prompts to guide the image synthesis process by leveraging the available contextual information in a few text examples. The results revealed that synthesized images using the proposed prompt optimization method achieved a higher performance, at 18% on average, based on an evaluation metric that measures the similarity between the generated images and prompts for generation. The significance of this research lies in its potential to provide a more efficient and automated approach to obtaining high-quality synthesized images. The findings indicate that prompt optimization may offer a promising pathway for text-to-image generative models. | Lee, Seunghun; Lee, Jihoon; Bae, Chan Ho; Choi, Myung-Seok; Lee, Ryong; Ahn, Sangtae | Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu 41566, South Korea; Korea Inst Sci & Technol Informat KISTI, AI Data Res Ctr, Daejeon 34141, South Korea; Kyungpook Natl Univ, Sch Elect Engn, Daegu 41566, South Korea | ; Ahn, Sangtae/AFQ-7342-2022 | 58847507500; 59448690600; 58795923800; 55461985900; 34880307100; 55468016100 | ryonglee@kisti.re.kr;stahn@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 2.58 | 2025-04-16 | 4 | 6 | Tuning; Computational modeling; Training; Task analysis; Data models; Image synthesis; Visualization; Text categorization; In-context few-shot learning; pretrained language model; prompt optimization; text-to-image generation | In-context few-shot learning; pretrained language model; prompt optimization; text-to-image generation | Computational linguistics; Job analysis; Modeling languages; Text processing; Computational modelling; Image generations; Images synthesis; In contexts; In-context few-shot learning; Language model; Optimisations; Pretrained language model; Prompt optimization; Task analysis; Text categorization; Text-to-image generation; Tuning; Image processing | English | 2024 | 2024 | 10.1109/access.2023.3348778 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | Phase Optimized Computer-Generated Holographic Video Calculation With Frame Interpolation Using Gradient Descent Algorithm | Computer-generated holography (CGH) has been anticipated in augmented reality (AR) field since it can fully provide multi-depth 3D information to users. As the gradient descent algorithms have been developed and intensively applied to the holography, traditional challenges in CGH, such as speckle noises and excessive computation load, have been overcome for a given object. However, despite the widespread consumption form of media contents, studies for the frame interpolation of high-quality computer-generated holographic video (CGHV) has not been substantially conducted yet. Here, we demonstrate a method for rapid calculation of speckle-free, frame interpolated CGHV using stochastic gradient descent algorithm, both in numerical and experimental results. We demonstrate that the similarity between input and target images is related to a speed of convergence to desired image quality levels, and our method enables the generation of interpolated frames with controlled intensity ratios during the optimization process. Our proposed method can reduce the burden of computation for high-quality CGHV, so that it can be potentially used for holographic display and AR applications such as AR head-up display in automobile. | Jin, Gyeongsu; Ann, Young-Jun; Lee, Seung-Yeol | Kyungpook Natl Univ, Coll IT Engn, Sch Elect & Elect Engn, Daegu 41566, South Korea | 59405483900; 58242643700; 55881869300 | seungyeol@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0.39 | 2025-05-07 | 1 | 1 | Holography; computer-generated hologram; stochastic gradient descent; frame interpolation; Holography; computer-generated hologram; stochastic gradient descent; frame interpolation | FULL-RANGE | computer-generated hologram; frame interpolation; Holography; stochastic gradient descent | Augmented reality; Automobile frames; Computer generated holography; Electron holography; Gradient methods; Helmet mounted displays; Holograms; Holographic displays; Image quality; Interpolation; Video recording; 3D information; Computer generated; Computer-generated holography; Computergenerated holograms (CGH); Frame interpolation; Gradient descent algorithms; High quality; Holographic videos; Speckle noise; Stochastic gradient descent; Head-up displays | English | 2024 | 2024 | 10.1109/access.2024.3491223 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | Priority-Encoder Ensemble for Speech Recognition | The advancement in computational capabilities and the availability of vast datasets have propelled the performance of Automatic Speech Recognition (ASR) systems. However, the task of ASR is complex, requiring consideration of diverse factors such as spoken tone, intonation, accents, and pitch modulation. To tackle these challenges, ensembles of Large Language Models (LLMs) have emerged as a promising approach, harnessing the strengths of multiple models to improve recognition accuracy. These ensembles, employing various strategies, often encounter significant time requirements during the inference process limiting the applicability in real-life scenarios. In this study, we introduce a novel ensemble strategy, the Priority-Encoder Ensemble (PE-Ensemble), for ASR systems. The PE-Ensemble employs a meta-learning-based Decider model to dynamically select the optimal model from the ensemble for inference, significantly reducing the computational load and memory requirements during inference. Unlike traditional ensembles where all models are loaded into memory, our approach requires only a single model to be loaded, enhancing efficiency in real-world applications such as unmanned kiosks. We evaluate the PE-Ensemble against the commonly used average ensemble strategy and individual base models. The results demonstrate that the PE-Ensemble outperforms both the average ensemble and individual base models in terms of prediction accuracy as well as computational time during inference. This enhancement in accuracy, coupled with the substantial reduction in computational load, highlights the efficacy and practical applicability of the proposed PE-Ensemble approach. | Ivan, Dzeuban Fenyom; Darlan, Daison; Adedigba, Adeyinka; Ajani, Oladayo S.; Mallipeddi, Rammohan; Joo, Hwang Jae | Kyungpook Natl Univ, Sch Elect Engn, Dept Artificial Intelligence, Daegu 37224, South Korea; Kyungpook Natl Univ, Smart Agr Innovat Ctr, Daegu 41566, South Korea; Gwang Myeong Tech Co Ltd, Daegu 41566, South Korea | ; Adedigba, Adeyinka/KIC-7325-2024; AJANI, Oladayo/HIR-9607-2022; Mallipeddi, Rammohan/AAL-5306-2020; Darlan, Daison/KQA-9542-2024 | 58655933000; 58164208500; 57194027040; 57465126000; 25639919900; 59323124100 | mallipeddi.ram@gmail.com; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0 | 2025-05-07 | 0 | 0 | Predictive models; Computational modeling; Load modeling; Adaptation models; Training; Data models; Ensemble learning; Automatic speech recognition; Encoding; priority-encoder; ensemble of models | CLASSIFIER | Automatic speech recognition; ensemble of models; priority-encoder | Audio signal processing; Signal encoding; Signal modulation; Automatic speech recognition; Automatic speech recognition system; Base models; Computational capability; Computational loads; Ensemble of models; Ensemble strategies; Multiple-modeling; Performance; Priority Encoder; Speech recognition | English | 2024 | 2024 | 10.1109/access.2024.3454221 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||
| ○ | ○ | Article | RandMixAugment: A Novel Unified Technique for Region- and Image-Level Data Augmentations | Deep learning models learn powerful representational spaces required for handling complex tasks. Recently, data augmentation techniques, region-level, and image-level augmentation have proved effective in significantly improving deep learning models' generalization performance. Nevertheless, such methods may lose some critical features or are still computationally heavy (or inefficient) due to additional computation burdens. To address this issue, in this paper, we present a novel unified data augmentation method for deep learning models, namely, RandMixAugment, which effectively combines the intrinsic properties of region-level augmentation and image-level augmentation. Specifically, the proposed RandMixAugment employs automated augmentation with masking and mixing operations. Experiments are conducted on well-known CIFAR datasets (CIFAR-10 and CIFAR-100) to verify the effectiveness of the proposed scheme compared to state-of-the-art augmentation techniques. The experimental results demonstrate that the proposed RandMixAugment yields superior performance over state-of-the-art techniques on image classification tasks and further improves the performance of the baseline deep learning model by 1.2% and 2.4% on CIFAR-10 and CIFAR-100 datasets, respectively. | Shin, Yosoeb; Palakonda, Vikas; Yun, Sangseok; Kim, Il-Min; Kim, Seon-Gon; Park, Sang-Mi; Kang, Jae-Mo | Kyungpook Natl Univ, Dept Artificial Intelligence, Daegu 41566, South Korea; Pukyong Natl Univ, Dept Informat & Commun Engn, Busan 48513, South Korea; Queens Univ, Dept Elect & Comp Engn, Kingston, ON K7L 3N6, Canada; KORAIL Res Safety Inst, Technol Res Safety Dept, Daejeon 34618, South Korea | 58540649500; 57193028485; 56115729600; 36040390300; 58540292900; 58540293000; 56024930400 | jmkang@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0.86 | 2025-04-16 | 2 | 2 | Classification; data augmentation; deep learning; image processing; supervised learning | Classification; data augmentation; deep learning; image processing; supervised learning | Classification (of information); Deep learning; Image classification; Image enhancement; Job analysis; Personnel training; Classification algorithm; Computational modelling; Data augmentation; Deep learning; Images classification; Images processing; Task analysis; Supervised learning | English | 2024 | 2024 | 10.1109/access.2023.3305385 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||||
| ○ | ○ | Article | Real-Time Reinforcement Learning for Optimal Viewpoint Selection in Monocular 3D Human Pose Estimation | Monocular 3D human pose estimation (HPE) presents an inherently ill-posed challenge, complicated by issues such as depth ambiguity and uncertainty. Estimating 3D poses with a single camera heavily depends on viewpoint, resulting in poor pose estimation accuracy. To address these challenges, we propose a real-time reinforcement learning-based viewpoint selection method that dynamically adjusts the camera viewpoint to optimize pose estimation. Our method extracts features encoding depth ambiguity and uncertainty from 2D-to-3D lifting, allowing the model to identify the optimal camera movements without requiring multiple cameras. We evaluate our approach on a publicly available real-world dataset, adjusted to simulate a realistic setting of drone flights capturing human motions. Our approach, compared against baseline strategies including fixed, random, and rotating camera movements with various 3D HPE models, significantly enhances the accuracy and robustness of pose estimation. In particular, it achieves a notable improvement, reducing pose estimation errors by over 30% compared to fixed and random camera movements. These results highlight the effectiveness of our method in optimizing viewpoint selection for real-time 3D HPE, making it a practical solution for single-camera setups in dynamic environments. Our code is available at https://github.com/knu-vis/nbv-pose. | Lee, Sanghyeon; Hwang, Yoonho; Lee, Jong Taek | Kyungpook Natl Univ, Sch Comp Sci & Engn, Daegu 41566, South Korea | 57226171120; 59483129100; 59484914800 | jongtaeklee@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0 | 2025-05-07 | 0 | 0 | Three-dimensional displays; Cameras; Real-time systems; Accuracy; Pose estimation; Heating systems; Uncertainty; Drones; Solid modeling; Feature extraction; 3D human pose estimation; next best viewpoint selection; deep learning; reinforcement learning | 3D human pose estimation; deep learning; next best viewpoint selection; reinforcement learning | Adversarial machine learning; Contrastive Learning; Deep learning; Federated learning; Reinforcement learning; 3D human pose estimation; Camera's movements; Deep learning; Next best viewpoint selection; Next-best viewpoint; Pose-estimation; Real- time; Reinforcement learnings; Uncertainty; Viewpoint selection; Deep reinforcement learning | English | 2024 | 2024 | 10.1109/access.2024.3514146 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | |||||
| ○ | ○ | Article | Short-Term Fault Prediction of Wind Turbines Based on Integrated RNN-LSTM | This paper presents a data-driven approach to short-term wind turbine fault prediction and condition monitoring based on a hybrid architecture of recurrent neural network and long short-term memory. The proposed architecture is established by utilizing time series data from the supervisory control and data acquisition system and a Bladed model of a 5 MW wind turbine to predict faults occurring to the wind generator. The recurrent neural network-long short-term memory training procedure is enhanced with self-organizing maps and long short-term memory auto encoder so as to describe the complex interaction between the mechanical system and unpredictable wind speed. To verify the performance of the proposed scheme, we conduct in-depth numerical experiments by applying the hybrid architecture to the Bladed 5 MW wind turbine model with rated wind speed of 11.8 m/s. Experimental results confirm that the proposed scheme has superior accuracy and practicality of fault prediction compared with eminent existing machine learning algorithms such as extreme gradient boost and random forest regressor. | Rama, V. Siva Brahmaiah; Hur, Sung-Ho; Yang, Jung-Min | Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu 41566, South Korea | 57437775600; 36455858700; 57208450551 | shur@knu.ac.kr;jmyang@ee.knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 4.72 | 2025-04-16 | 10 | 14 | Extreme gradient boosting; fault prediction; long short-term memory autoencoder; random forest regressor; self-organizing maps; wind turbine | ANOMALY DETECTION; MAINTENANCE; OPERATION; TRENDS | Extreme gradient boosting; fault prediction; long short-term memory autoencoder; random forest regressor; self-organizing maps; wind turbine | Brain; Condition monitoring; Conformal mapping; Fault detection; Forecasting; Learning algorithms; Learning systems; Long short-term memory; Memory architecture; Network architecture; Self organizing maps; Signal encoding; Time series analysis; Wind speed; Auto encoders; Computational modelling; Extreme gradient boosting (xgboost); Fault prediction; Faults detection; Gradient boosting; Long short-term memory auto encoder; Predictive models; Random forest regressor; Random forests; Self-organizing map; Self-organizing-maps; Time-series analysis; Wind speed; Wind turbines | English | 2024 | 2024 | 10.1109/access.2024.3364395 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | ○ | Article | TDO-SLAM: Traffic Sign and Dynamic Object Based Visual SLAM | This paper introduces a real-time visual SLAM system, TDO-SLAM, using only a stereo vision camera. TDO-SLAM works not only in static but also in dynamic road environment by incorporating the object motion and the planar property of standing traffic signs. Traditional visual SLAM systems assume that the road environment is static. However, a variety of dynamic objects exist in the real-world urban environment. Thus, the traditional SLAM systems are subject to fail due to the various motion of the dynamic objects. To solve this inherent problem in the dynamic environment, TDO-SLAM detects, tracks, and manages the global object identification of dynamic objects and standing traffic signs through a novel Object-Level-Tracking method. We improve the accuracy of camera pose estimation through several steps of bundle adjustments, including the residual terms for the planar constraint of traffic signs and the dynamic object motion. Experimental results show that pose estimation accuracy is improved in complex environment with several dynamic objects and traffic signs. Performance of TDO-SLAM is analyzed and compared with ORB-SLAM2, ORB-SLAM3, and DynaSLAM using three benchmark datasets, KITTI Odometry dataset, KITTI Raw dataset, and Complex Urban dataset. | Park, Soon-Yong; Lee, Junesuk | Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu 41566, South Korea; 42dot, Seoul 06267, South Korea | ; Park, Soon-Yong/HGV-2374-2022 | 7501834063; 57210786619 | sypark@knu.ac.kr; | IEEE ACCESS | IEEE ACCESS | 2169-3536 | 12 | SCIE | ENGINEERING, ELECTRICAL & ELECTRONIC;TELECOMMUNICATIONS;COMPUTER SCIENCE, INFORMATION SYSTEMS | 2024 | 3.6 | 34.8 | 0.43 | 2025-04-16 | 1 | 2 | Dynamic SLAM; visual localization; pose estimation; autonomous vehicle | ROBOTICS; TRACKING | autonomous vehicle; Dynamic SLAM; pose estimation; visual localization | Benchmarking; Cameras; Computer vision; Feature extraction; Roads and streets; Robotics; Stereo image processing; Stereo vision; Autonomous Vehicles; Dynamic objects; Dynamic SLAM; Features extraction; Pose-estimation; Simultaneous localization and mapping; Vehicle's dynamics; Visual localization; Visual SLAM; Traffic signs | English | 2024 | 2024 | 10.1109/access.2024.3362675 | 바로가기 | 바로가기 | 바로가기 | 바로가기 |
페이지 이동: