연구성과로 돌아가기
2023 연구성과 (285 / 285)
※ 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 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| ○ | ○ | Proceedings Paper | Vision Transformer Compression and Architecture Exploration with Efficient Embedding Space Search | This paper addresses theoretical and practical problems in the compression of vision transformers for resource-constrained environments. We found that deep feature collapse and gradient collapse can occur during the search process for the vision transformer compression. Deep feature collapse diminishes feature diversity rapidly as the layer depth deepens, and gradient collapse causes gradient explosion in training. Against these issues, we propose a novel framework, called VTCA, for accomplishing vision transformer compression and architecture exploration jointly with embedding space search using Bayesian optimization. In this framework, we formulate block-wise removal, shrinkage, cross-block skip augmentation to prevent deep feature collapse, and Res-Post layer normalization to prevent gradient collapse under a knowledge distillation loss. In the search phase, we adopt a training speed estimation for a large-scale dataset and propose a novel elastic reward function that can represent a generalized manifold of rewards. Experiments were conducted with DeiT-Tiny/Small/Base backbones on the ImageNet, and our approach achieved competitive accuracy to recent patch reduction and pruning methods. The code is available at https://github.com/kdaeho27/VTCA. | Kim, Daeho; Kim, Jaeil | Kyungpook Natl Univ, Dept Artificial Intelligence, Daegu, South Korea; Kyungpook Natl Univ, Sch Comp Sci & Engn, Daegu, South Korea | Kim, Dae-ho/AAY-7919-2020 | 57216887648; 57211615348 | threeyears@gmail.com; | COMPUTER VISION - ACCV 2022, PT III | 0302-9743 | 1611-3349 | 13843 | 0.52 | 2025-06-25 | 1 | 1 | Computer vision; Distillation; Large dataset; Architecture exploration; Bayesian optimization; Embeddings; Normalisation; Practical problems; Search phasis; Search process; Space search; Speed estimation; Training speed; Embeddings | English | 2023 | 2023 | 10.1007/978-3-031-26313-2_32 | 바로가기 | 바로가기 | 바로가기 | ||||||||||||
| ○ | ○ | Proceedings Paper | Visual LiDAR Odometry Using Tree Trunk Detection and LiDAR Localization | This paper presents a method of visual LiDAR odometry and forest mapping, leveraging tree trunk detection and LiDAR localization techniques. In environments like dense forests, where smooth GPS signals are unreliable, we employ camera and LiDAR sensors to accurately estimate the robot's position. However, forested or orchard settings introduce unique challenges, including a diverse mixture of trees, tall grass, and uneven terrain. To address these complexities, we propose a distance-based filtering method to extract data composed solely of tree trunk information from 2D LiDAR. By restoring arc data from the LiDAR sensor to its circular shape, we obtain position and radius measurements of reference trees in the LiDAR coordinate system. Then, these values are stored in a comprehensive tree trunk database. Our approach combines visual-based SLAM and LiDAR-based SLAM independently, followed by an integration step using the Extended Kalman Filter (EKF) to improve odometry estimation. Utilizing the obtained odometry information and the EKF, we generate a tree map based on observed trees. In addition, we use the tree position in the map as the landmark to reduce the localization error in the proposed SLAM algorithm. Experimental results show that the loop-closing error ranges between 0.3 to 0.5 meters. In the future, it is expected that this method will also be applicable in the fields of path planning and navigation. | Park, K. W.; Park, S. Y. | Kyungpook Natl Univ, Grad Sch Elect & Elect Engn, Daegu 41566, South Korea | Park, Soon-Yong/HGV-2374-2022 | 58846778200; 7501834063 | kevin2760@naver.com;sypark@knu.ac.kr; | GEOSPATIAL WEEK 2023, VOL. 48-1 | 1682-1750 | 2194-9034 | 0.92 | 2025-06-25 | 1 | 2 | SLAM; Tree Trunk; Mapping; Stereo Camera; LiDAR | LiDAR; Mapping; SLAM; Stereo Camera; Tree Trunk | Cameras; Extended Kalman filters; Forestry; Information filtering; Motion planning; Optical radar; Robots; Stereo image processing; Forest mapping; GPS signals; LiDAR; Localisation; Localization technique; Odometry; Robot positions; SLAM; Stereo cameras; Tree trunk; Mapping | English | 2023 | 2023 | 10.5194/isprs-archives-xlviii-1-w2-2023-627-2023 | 바로가기 | 바로가기 | 바로가기 | |||||||||||
| ○ | Conference paper | Visual Tactile Sensor based on Feature Tracking of Patterns for Soft Human-Machine Interaction | Tactile sensors are used in various fields such as automated factories and human collaboration. Tactile sensors exist in a variety of technological ways. In particular, most of the contact determination methods are mainly performed through the detection of the physical surface. In contrast, recently, various Vision-based tactile sensors that can replace the existing method based only on visual data from an image viewed through a camera sensor have been proposed. The hardware proposed in this paper is also a Vision-based tactile sensor, and it is a method that determines contact based only on patterns. In addition, we propose a vision-based tactile sensor as hardware in the form of air bag based on an air cushion. As the biggest feature of the Vision-based tactile sensor is estimation through image reading, it is easy to update various functions through algorithm improvement. Based on these points, through continuous research, we will develop algorithms for position estimation stability improvement, force estimation, and multi-touch discrimination, among at the possibility of application to fields such as cooperative human interaction robots. © 2023 IEEE. | Lee, Jinhyuk; Lee, Suwoong; Kim, Min Young | Kyungpook National University, School of Electronic and Electrical Engineering, Daegu, 41566, South Korea; Technology Congergence Group, Korea Institude of Industrial Technology (KITECH), Department of Mechatronics, Daegu, 42994, South Korea; Kyungpook National University, School of Electronics Engineering, Daegu, 41566, South Korea | 58072935700; 57188756166; 56739349100 | 0920wlsgur@naver.com;minykim@knu.ac.kr; | International Conference on Ubiquitous and Future Networks, ICUFN | 2165-8528 | 2023-July | 0 | 2025-06-25 | 0 | air cushion; contact estimation; image processing; Vision-based Tactile Sensor | Human robot interaction; Tactile sensors; Air cushion; Contact estimation; Determination methods; Feature-tracking; Human machine interaction; Images processing; Tactile sensors; Vision based; Vision-based tactile sensor; Visual data; Image enhancement | English | Final | 2023 | 10.1109/icufn57995.2023.10201138 | 바로가기 | 바로가기 | ||||||||||||||||
| ○ | ○ | Proceedings Paper | Volumetric Body Composition Through Cross-Domain Consistency Training for Unsupervised Domain Adaptation | Computed tomography (CT) scans of the abdomen have emerged as a robust, precise, and dependable means of determining body composition. The accurate prediction of skeletal muscle volume (SMV) using slices of CT scans holds critical importance in facilitating subsequent diagnosis and prognosis. A significant proportion of research in the field of abdominal image analysis is primarily focused on the third lumbar spine vertebra (L3), owing to two prominent factors. Firstly, L3 is a large vertebra situated in the middle of the lumbar spine, rendering it less susceptible to degenerative changes in comparison to other lumbar vertebrae, making it a stable landmark. Secondly, the slice labeling in a CT volume is an intricate and time-consuming process, demanding significant human efforts, whereas labeling a single slice from a specific vertebral level is comparatively simpler. This study leverages labeled L3 slices i.e., source domain to reliably predict unlabeled lumbar region slices other than L3 i.e., target domain. We use Cross-Domain Consistency Training (CDCT) to extend network's current knowledge, acquired through segmenting a source domain, by learning to label a target domain. A consistency is enforced between the predictions from two segmentation networks with identical lightweight architecture but have different weight initialization points. The training objective consists of supervised loss terms for the source domain data and unsupervised loss terms for the target domain data. Remarkably, our trained network exhibits a marked enhancement in performance when applied to the target domain, indicating domain invariant feature learning through cross-domain consistency training could significantly enhance a network's generalization capability. | Ali, Shahzad; Lee, Yu Rim; Park, Soo Young; Tak, Won Young; Jung, Soon Ki | Kyungpook Natl Univ, Sch Comp Sci & Engn, Daegu, South Korea; Kyungpook Natl Univ, Kyungpook Natl Univ Hosp, Dept Internal Med, Coll Med, Daegu, South Korea | ; Ali, Shahzad/GPG-6925-2022; Lee, YuRim/GRF-4873-2022; Jung, Soon Ki/P-7687-2018 | 57709386500; 57194094753; 57191674344; 7004074582; 57226791905 | shahzadali@knu.ac.kr;skjung@knu.ac.kr; | ADVANCES IN VISUAL COMPUTING, ISVC 2023, PT I | 0302-9743 | 1611-3349 | 14361 | 1.03 | 2025-06-25 | 2 | 2 | Volumetric Body Composition; Skeletal Muscle Volume; Unsupervised Domain Adaptation; Abdominal CT Segmentation | Abdominal CT Segmentation; Skeletal Muscle Volume; Unsupervised Domain Adaptation; Volumetric Body Composition | Biochemistry; Computerized tomography; Image segmentation; Muscle; Abdominal computed tomographies; Abdominal computed tomography segmentation; Body composition; Domain adaptation; Muscle volume; Skeletal muscle; Skeletal muscle volume; Unsupervised domain adaptation; Volumetric body composition; Volumetrics; Forecasting | English | 2023 | 2023 | 10.1007/978-3-031-47969-4_23 | 바로가기 | 바로가기 | 바로가기 | ||||||||||
| ○ | ○ | Article | Volunteer Plants' Occurrence and the Environmental Adaptability of Genetically Modified Fodder Corn upon Unintentional Release into the Environment | The number of corn cultivars that have been improved using genetically modified technology continues to increase. However, concerns about the unintentional release of living-modified organisms (LMOs) into the environment still exist. Specifically, there are cases where LMO crops grown as fodder are released into the environment and form a volunteer plant community, which raises concerns about their safety. In this study, we analyzed the possibility of weediness and volunteer plants' occurrence when GMO fodder corn grains distributed in Korea are unintentionally released into the environment. Volunteer plants' occurrence was investigated by directly sowing grains in an untreated field. The results showed that the germination rate was extremely low, and even if a corn seed germinated, it could not grow into an adult plant and would die due to weed competition. In addition, the germination rate of edible and fodder grains was affected by temperature (it was high at 20 & DEG;C and 30 & DEG;C but low at 40 & DEG;C and extremely low at 10 & DEG;C), and it was higher in the former than in the latter. And the germination rate was higher in Daehakchal (edible corn grains) than in Gwangpyeongok (fodder corn grains). The environmental risk assessment data obtained in this study can be used for future evaluations of the weediness potential of crops and the development of volunteer plant suppression technology in response to unintentional GMO release. | Choi, Han-Yong; Kim, Eun-Gyeong; Park, Jae-Ryoung; Jang, Yoon-Hee; Jan, Rahmatullah; Farooq, Muhammad; Asif, Saleem; Kim, Nari; Kim, Ji-Hun; Gwon, Dohyeong; Lee, Seong-Beom; Jeong, Seung-Kyo; Kim, Kyung-Min | Kyungpook Natl Univ, Dept Appl Biosci, Daegu 41566, South Korea; Kyungpook Natl Univ, Coastal Agr Res Inst, Daegu 41566, South Korea; Rural Dev Adm, Natl Inst Crop Sci, Crop Breeding Div, Wonju 55365, South Korea | ; Jan, Rahmatullah/AIC-3439-2022; Kim, Kyung-Min Kim/C-7007-2014 | 58513516900; 57221496070; 57211205505; 57219901992; 57201981969; 57215544380; 57396413700; 57395985700; 56024681400; 58512306900; 58513030200; 58512307000; 34868260300 | seedsaler@naver.com;egk@knu.ac.kr;icd0192@korea.kr;uni@knu.ac.kr;rehmatbot@yahoo.com;mfarooqsr@gmail.com;saleemasif10@gmail.com;jennynari@hanmail.net;kim960520@naver.com;pondelionn@naver.com;dltks6954@naver.com;tmdry1221@gmail.com;kkm@knu.ac.kr; | PLANTS-BASEL | 2223-7747 | 12 | 14 | 0 | 2025-06-25 | 0 | 0 | GMO; corn; release; volunteer plants; weediness | FOOD; IMPACTS; MAIZE; CLIMATE; CHINA; FEED | corn; GMO; release; volunteer plants; weediness | English | 2023 | 2023-07 | 10.3390/plants12142653 | 바로가기 | 바로가기 | 바로가기 | ||||||||||
| ○ | Article | Wave Simulation Technique for Large-scale Optical Sensor Designs | The wave mode calculation of a large-scale optical system in comparison to the working wavelength is practically impossible because the computational cost increases exponentially. In this paper, we propose a method that can obtain the optical mode in a large-scale optical system. The method carries out simulations by dividing the calculation area into blocks and moving along the light axis along which the light propagates. By applying this method to the calculation of resonant modes in a ring-type optical resonator, which is mainly used for ring laser optical gyro sensors, the efficiency of the proposed method was verified. © 2023, Korean Sensors Society. All rights reserved. | Lee, Yong-Hoon; Kwon, Tae Yoon; Choi, Muhan | School of Electronics Engineering, Kyungpook National University, 80, Daehak-ro, Buk-gu, Daegu, 41566, South Korea; Department of Navigation Systems, Hanwha Munitions Corporation, 10, Yuseong-daero 1366beon-gil, Yuseong-gu, Daejeon, 34101, South Korea; School of Electronics Engineering, Kyungpook National University, 80, Daehak-ro, Buk-gu, Daegu, 41566, South Korea | 57219622064; 24481593600; 7402093793 | mhchoi@ee.knu.ac.kr; | Journal of Sensor Science and Technology | 1225-5475 | 32 | 1 | 0.15 | 2025-06-25 | 2 | Finite difference time domain (FDTD) method; Gyro-sensor; Resonant mode; Ring resonator | Korean | Final | 2023 | 10.46670/jsst.2023.32.1.62 | 바로가기 | 바로가기 | ||||||||||||||||
| ○ | Article | Whole-genome sequence of priestia aryabhattai strain s2 isolated from the rhizosphere of soybean (Glycine max) | We present the complete genome sequence of Priestia aryabhattai strain S2 isolated from the soybean rhizosphere. The genome consists of a single circular chromosome of 5,070,860 bp with a G+C content of 38.3% and 2 plasmids, P1(148,124 bp, GC content 33.3%) and P2 (76,418 bp, GC content 36.5%). © 2023, The Korean Society for Microbiology and Biotechnology. | Sliti, Amani; Kim, Min-Ji; Lee, Gyudae; Yeong-Junpark; Shin, Jae-Ho | Department of Applied Biosciences, Kyungpook National University, Daegu, 41566, South Korea; Department of Applied Biosciences, Kyungpook National University, Daegu, 41566, South Korea; Department of Applied Biosciences, Kyungpook National University, Daegu, 41566, South Korea; NGS Core Facility, Kyungpook National University, Daegu, 41566, South Korea; Department of Applied Biosciences, Kyungpook National University, Daegu, 41566, South Korea, NGS Core Facility, Kyungpook National University, Daegu, 41566, South Korea | 58551490600; 57127351600; 57222101785; 58641399000; 57224125922 | jhshin@knu.ackr; | Microbiology and Biotechnology Letters | 1598-642X | 51 | 3 | 0 | 2025-06-25 | 0 | Genome; Priestia aryabhattai; Rhizosphere; Soybean | article; chromosome 5; DNA base composition; nonhuman; plasmid; rhizosphere; soybean | English | Final | 2023 | 10.48022/mbl.2307.07010 | 바로가기 | 바로가기 | |||||||||||||||
| ○ | Conference paper | Wide viewing angle holographic augmented reality near-eye display with holographic optical element | Holographic Optical Elements (HOEs) have emerged as a pivotal technology in enhancing holographic Augmented Reality (AR) display systems. This paper presents an innovative approach that utilizes an off-axis arrangement with an HOE to secure a wide field of view up to 55°, making significant strides in volumetric reduction compared to conventional 4f filtering systems. However, a challenge arises from Bragg mismatch in the HOE, which creates aberrations. Our work proposes a method for compensating these aberrations on a voxel-by-voxel basis, substantially improving the quality of the holographic display. Limitations such as the 2mm maximum size of the eye box due to the diffraction limit of the spatial light modulator (SLM) are acknowledged, but we suggest potential solutions such as using the HOE substrate glass as a waveguide and incorporating an array of lenses with an eye tracker for pupil tracking. Our findings offer significant contributions to the holographic display technology landscape and suggest promising avenues for future research. © 2023 SPIE. | Lee, Seongju; Jeon, Hosung; Moon, Woonchan; Jung, Minwoo; Hahn, Joonku | School of Electronic and Electrical Engineering, Kyungpook National University, 80 Daehak-ro, Buk-gu, Daegu, 41566, South Korea; School of Electronic and Electrical Engineering, Kyungpook National University, 80 Daehak-ro, Buk-gu, Daegu, 41566, South Korea; School of Electronic and Electrical Engineering, Kyungpook National University, 80 Daehak-ro, Buk-gu, Daegu, 41566, South Korea; School of Electronic and Electrical Engineering, Kyungpook National University, 80 Daehak-ro, Buk-gu, Daegu, 41566, South Korea; School of Electronic and Electrical Engineering, Kyungpook National University, 80 Daehak-ro, Buk-gu, Daegu, 41566, South Korea | 58636781900; 56663784700; 56340460500; 57216159562; 10142501600 | jhahn@knu.ac.kr; | Proceedings of SPIE - The International Society for Optical Engineering | 0277-786X | 12624 | 0 | 2025-06-25 | 0 | aberration compensation; augmented reality; Hologram; holographic optical element; near-eye display; optical design | Aberrations; Augmented reality; Diffraction; Eye tracking; Holographic displays; Holographic optical elements; Light modulators; Optical data processing; Optical design; % reductions; Aberration compensation; Display system; Filtering systems; Innovative approaches; Near-eye display; Off-axis; Volumetrics; Wide field-ofview; Wide viewing angle; Holograms | English | Final | 2023 | 10.1117/12.2675822 | 바로가기 | 바로가기 | ||||||||||||||||
| ○ | Proceedings Paper | WIND TURBINE GUST CONTROL USING LIDAR-ASSISTED MODEL PREDICTIVE CONTROL | Light Detection and Ranging (LIDAR)-based wind measurement system, positioned forward-facing, can gather information about the approaching wind. It proactively enables the wind turbine to adjust its operation via the feedforward (FF) loop. LIDAR technology can enhance wind turbine performance throughout its entire operational range. It can assist in torque control when wind speeds are below the rated level and in pitch control when wind speeds exceed the rated level. In this study, Model Predictive Control (MPC) is utilized. Within the field of wind turbine research, MPC has garnered significant interest in recent years due to its capability to handle both input and output constraints and leverage advanced on disturbances caused by the incoming wind, measured by the LIDAR. In this study, an FF-MPC is designed and compared with a more standard feedback (FB) MPC. A comparison is conducted in realistic gust wind conditions, considering below and above-rated wind ranges. These comparisons are performed in a realistic, high-fidelity aeroelastic simulation environment, i.e., DNV BLADED. Both controllers are designed for the DNV BLADED Supergen 5 MW wind turbine model. The control algorithm is implemented in C++, compiled into a dynamic link library (DLL), and integrated as an external controller within the DNV BLADED to enable accurate, high-fidelity simulations. Simulation results are presented to demonstrate the superiority of FF-MPC over the standard FB-MPC. | Reddy, Yiza Srikanth; Hur, Sung-ho | Kyungpook Natl Univ, Sch Elect & Elect Engn, Daegu, South Korea | shur@knu.ac.kr; | PROCEEDINGS OF ASME 2023 5TH INTERNATIONAL OFFSHORE WIND TECHNICAL CONFERENCE, IOWTC2023 | 1 | Feedforward control; model predictive control; LIDAR-assisted control | FEEDFORWARD CONTROL; PITCH CONTROL; DESIGN | English | 2023 | 2023 | 바로가기 | |||||||||||||||||||||||
| ○ | Conference paper | WIND TURBINE GUST CONTROL USING LIDAR-ASSISTED MODEL PREDICTIVE CONTROL | Light Detection and Ranging (LIDAR)-based wind measurement system, positioned forward-facing, can gather information about the approaching wind. It proactively enables the wind turbine to adjust its operation via the feedforward (FF) loop. LIDAR technology can enhance wind turbine performance throughout its entire operational range. It can assist in torque control when wind speeds are below the rated level and in pitch control when wind speeds exceed the rated level. In this study, Model Predictive Control (MPC) is utilized. Within the field of wind turbine research, MPC has garnered significant interest in recent years due to its capability to handle both input and output constraints and leverage advanced on disturbances caused by the incoming wind, measured by the LIDAR. In this study, an FF-MPC is designed and compared with a more standard feedback (FB) MPC. A comparison is conducted in realistic gust wind conditions, considering below and above-rated wind ranges. These comparisons are performed in a realistic, high-fidelity aeroelastic simulation environment, i.e., DNV BLADED. Both controllers are designed for the DNV BLADED Supergen 5 MW wind turbine model. The control algorithm is implemented in C++, compiled into a dynamic link library (DLL), and integrated as an external controller within the DNV BLADED to enable accurate, high-fidelity simulations. Simulation results are presented to demonstrate the superiority of FF-MPC over the standard FB-MPC. Copyright © 2023 by ASME. | Reddy, Yiza Srikanth; Hur, Sung-Ho | School of Electronic and Electrical Engineering, Kyungpook National University, Daegu, South Korea; School of Electronic and Electrical Engineering, Kyungpook National University, Daegu, South Korea | 57225000837; 36455858700 | Proceedings of ASME 2023 5th International Offshore Wind Technical Conference, IOWTC 2023 | 0.99 | 2025-06-25 | 1 | Feedforward control; LIDAR-assisted control; model predictive control | C++ (programming language); Controllers; Feedforward control; Optical radar; Predictive control systems; Wind; Wind turbines; Feedback model; Feedforward loops; Feedforward model; Light detection and ranging; Light detection and ranging-assisted control; Model-predictive control; Operational range; Turbine performance; Wind measurement system; Wind speed; Model predictive control | English | Final | 2023 | 10.1115/iowtc2023-119579 | 바로가기 | 바로가기 | |||||||||||||||||||
| ○ | ○ | Article | Word2Vec-based efficient privacy-preserving shared representation learning for federated recommendation system in a cross-device setting | Recommendation systems have required centralized storage of user data, but due to privacy concerns, recent studies adopted federated learning (FL) that discloses intermediate statistics instead of raw data to build privacy-preserving federated recommendation systems. However, they suffer from inefficiencies in privacy-preserving mechanisms and inaccuracies in simple algorithms that ignore sequential information. This study proposes an extension of Word2Vec for a privacy-preserving federated sequential recommendation system (PPFSRS). This method exploits sequential information to generate contextual item representations for accurate recommendations while concealing privacy-sensitive features efficiently. Specifically, we mixed updates from negative samples to inhibit the direct leakage of purchased items from model updates. In addition, our method computes approximate model updates that can occur when sensitive features only belong to negative samples to prevent inference attacks. In experiments, we used benchmark datasets for recommendation and simulated highly distributed data such that each user stores historical data locally. While preserving privacy with reasonable complexity, the proposed method showed little degradation in recommendation performance compared to FL-based Word2Vec without privacy preserving mechanisms. Utilizing contextual item representations trained by our method from highly distributed data will be a practical starting point for PPFSRS in a cross-device setting. | Lee, Taek-Ho; Kim, Suhyeon; Lee, Junghye; Jun, Chi-Hyuck | Seoul Natl Univ, Technol Management Econ & Policy Program, 1 Gwanak Ro, Seoul 08826, South Korea; Seoul Natl Univ, Inst Engn Res, 1 Gwanak Ro, Seoul 08826, South Korea; Kyungpook Natl Univ, Grad Sch Data Sci, Daegu 41566, South Korea; Seoul Natl Univ, Technol Management Econ & Policy Program, 1 Gwanak Ro, Seoul 08826, South Korea; Seoul Natl Univ, Grad Sch Engn Practice, 1 Gwanak Ro, Seoul 08826, South Korea; Pohang Univ Sci & Technol POSTECH, Dept Ind & Management Engn, 77 Cheongam Ro, Pohang 37673, South Korea | Lee, Junghye/KUF-0668-2024; Jun, Chi-Hyuck/AAE-1695-2019 | 57194686484; 57216511251; 56055191300; 16245299200 | junghye@snu.ac.kr;chjun@postech.ac.kr; | INFORMATION SCIENCES | INFORM SCIENCES | 0020-0255 | 1872-6291 | 651 | SCIE | COMPUTER SCIENCE, INFORMATION SYSTEMS | 2023 | N/A | 0.78 | 2025-06-25 | 7 | 7 | Privacy-preservation; Sequential recommender system; Federated learning; Word2Vec | Federated learning; Privacy-preservation; Sequential recommender system; Word2Vec | Digital storage; Learning systems; Privacy-preserving techniques; Distributed data; Federated learning; Model updates; Negative samples; Privacy preservation; Privacy preserving; Sensitive features; Sequential information; Sequential recommende system; Word2vec; Recommender systems | English | 2023 | 2023-12 | 10.1016/j.ins.2023.119728 | 바로가기 | 바로가기 | 바로가기 | 바로가기 | ||||
| ○ | Conference paper | Work-in-Progress: Micro-Accelerator-in-the-Loop Framework for MCU Integrated Accelerator Peripheral Fast Prototyping | The resource constraints of MCU-based platforms limits their ability to utilize high-performance accelerators such as GPUs or servers, mainly due to insufficient resources for ML applications. Currently, solutions utilizing accelerators connected as peripherals to the on-chip bus of microcontroller units (MCUs) are being proposed. We define this approach as a Micro-Accelerator (MA). Due to the necessity of connecting the MA to the MCU core and the on-chip bus within the chip, conducting a iterative full system evaluation of the embedded software that drives the MA poses significant challenges. To address this challenge, we propose a framework that enables rapid prototyping of custom-designed MA and facilitates profiling of its acceleration performance. Experimental results evaluating the performance of the MA for two tiny machine learning (TinyML) applications within the proposed framework demonstrate a cycle latency reduction of 84.32% and 61.32% compared to a general machine learning framework, respectively. © 2023 ACM. | Kwon, Jisu; Park, Daejin | Kyungpook National University, South Korea; Kyungpook National University, South Korea | 57215531728; 55463943600 | boltanut@knu.ac.kr; | Proceedings - 2023 International Conference on Embedded Software, EMSOFT 2023 | 0 | 2025-06-25 | 0 | Computer aided design; Microcontrollers; Program processors; Acceleration performance; Fast prototyping; Machine learning applications; Microcontroller unit; On-chip bus; Performance; Rapid-prototyping; Resource Constraint; System evaluation; Unit-based; Machine learning | English | Final | 2023 | 10.1145/3607890.3608461 | 바로가기 | 바로가기 | |||||||||||||||||||
| ○ | Conference paper | Work-in-Progress: Searching Optimal Compiler Optimization Passes Sequence for Reducing Runtime Memory Profile using Ensemble Reinforcement Learning | The order in which compiler optimization passes are applied has a significant impact on program performance. However, widely used compiler optimization options use handpicked sets of optimization passes, optimized for specific benchmarks. In this paper, we propose an ensemble reinforcement learning (RL) model that optimizes LLVM transform passes sequence to reduce the runtime memory profile, which is an important consideration in resource-constrained embedded systems. We developed an LLVM intermediate representation (IR) analysis pass to extract static program features. The extracted features are processed with PCA for dimension reduction. We also generated datasets using a random program generator, and clustered them according to the PCA results of their extracted features. The ensemble RL model was trained on each clustered dataset. Experiments showed that the proposed model reduced 37% more memory profile than the standard optimization option. © 2023 ACM. | Chang, Juneseo; Park, Daejin | Seoul National University, Seoul, South Korea; Kyungpook National University, Daegu, South Korea | 57393160100; 55463943600 | Proceedings - 2023 International Conference on Embedded Software, EMSOFT 2023 | 0.32 | 2025-06-25 | 1 | Code optimization; embedded system; reinforcement learning | Constrained optimization; Embedded systems; Learning systems; Optimal systems; Program compilers; Code optimization; Compiler optimizations; Embedded-system; Intermediate representations; Optimisations; Program performance; Reinforcement learning models; Reinforcement learnings; Resource-constrained embedded systems; Runtimes; Reinforcement learning | English | Final | 2023 | 10.1145/3607890.3608460 | 바로가기 | 바로가기 | |||||||||||||||||||
| ○ | Conference paper | Yield Monitoring Service with Time Series Representation of Growth from Crop Images | Due to the development of ICT technology and the explosion of data, various technologies such as smart farms and smart greenhouses have been applied to the agricultural sector, helping farmers a lot. This paper aims to process and utilize crop image datasets collected periodically from a strawberry greenhouse cultivation environment to analyze and monitor crop growth and yields. In addition to the time-series data collected by measuring environmental variables such as temperature, humidity, light intensity, and carbon dioxide concentration, growth-related data are needed to monitor fruit growth and predict yield. Rail cameras were installed on the testbed to collect image data on strawberry every certain time in a specific section, and to construct a time-series image dataset by detecting objects for each stage of strawberry growth. Due to the characteristics of strawberries, we represented the crop images into growth monitoring data applied to the entire fruit of the testbed or crop clusters, not to each fruit. In addition, it can be useful for various use cases if the represented growth-related time series data and sensor data are properly used together. © 2023 IEEE. | Oh, Seungtaek; Kim, Sung Kyeom; Moon, Jaewon; Choi, Seungwook; Lee, Soonho; Kum, Seungwoo; Suh, Hyun Kwon | Korea Electronics Technology Institute, Information Media Research Center, Seoul, South Korea; Kyungpook National University, Department of Horticultural Science, Daegu, South Korea; Korea Electronics Technology Institute, Information Media Research Center, Seoul, South Korea; Naretrends Inc., Bucheon, South Korea; Daliworks Inc., Seoul, South Korea; Korea Electronics Technology Institute, Information Media Research Center, Seoul, South Korea; Sejong University, Department of Integrative Biological Science and Industry, Seoul, South Korea | 57414318900; 50262290200; 37041654800; 58630209800; 58876021400; 35113505800; 57207105123 | stoh@keti.re.kr; | International Conference on ICT Convergence | 2162-1233 | 0.43 | 2025-06-25 | 1 | Crop monitoring; Data preparation; Object detection; Precise farming; Smart Farm; Time-series data; Yield monitoring | Carbon dioxide; Crops; Cultivation; Farms; Greenhouses; Testbeds; Crop monitoring; Data preparation; Image datasets; Monitoring services; Objects detection; Precise farming; Smart farm; Time-series data; Times series; Yield monitoring; Object detection | English | Final | 2023 | 10.1109/ictc58733.2023.10393247 | 바로가기 | 바로가기 |
페이지 이동: