Daegu Gyeongbuk Institute of Science and Technology
DGIST Library Institutional RepositoryNot a member yet
12664 research outputs found
Sort by
Effects of Computer Mouse Lift-off Distance Settings in Mouse Lifting Action
No full textThis study investigates the effect of Lift-off Distance (LoD) on a computer mouse, which refers to the height at which a mouse sensor stops tracking. Although a low LoD is generally preferred to avoid unintended cursor movement in mouse lifting (=clutching), especially in first-person shooter games, it may increase tracking errors. We conducted a psychophysical experiment to measure the perceptible differences between different LoD settings, and we quantitatively measured unintended cursor movement and tracking errors at four levels of LoD while users performed mouse lifting. The results quantified the amount of the two types of errors, which revealed the trade-off between them in the varying levels of LoD. © 2024 Association for Computing Machinery. All rights reserved
Sub-Terahertz Imaging-Based Real-Time Non-Destructive Inspection System for Estimating Water Activity and Foreign Matter Depth in Seaweed
No full textAs the importance of hygiene and safety management in food manufacturing has been increasingly emphasized, research on non-destructive and non-contact inspection technologies has become more active. This study proposes a real-time and non-destructive food inspection system with sub-terahertz waves which penetrates non-conducting materials by using a frequency of 0.1 THz. The proposed system detects not only the presence of foreign matter, but also the degree of depth to which it is mixed in foods. In addition, the system estimates water activity levels, which serves as the basis for assessing the freshness of seaweed by analyzing the transmittance of signals within the sub-terahertz image. The system employs YOLOv8n, which is one of the newest lightweight object detection models. This lightweight model utilizes the feature pyramid network (FPN) to effectively detect objects of various sizes while maintaining a fast processing speed and high performance. In particular, to validate the performance in real manufacturing facilities, we implemented a hardware platform, which accurately inspects seaweed products while cooperating with a conveyor device moving at a speed of 45 cm/s. For the validation of the estimation performance against various water activities and the degree of depth of foreign matter, we gathered and annotated a total of 9659 sub-terahertz images and optimized the learning model. The final results show that the precision rate is 0.91, recall rate is 0.95, F1-score is 0.93, and mAP is 0.97, respectively. Overall, the proposed system demonstrates an excellent performance in the detection of foreign matter and in freshness estimation, and can be applied in several applications regarding food safety. © 2024 by the authors.TRUEsciescopu
협력 지능형 교통 시스템에서 딥러닝과 차량의 물리적 속성을 활용한 통신 데이터 공격 탐지
No full textData falsification attack detection, Cooperative Intelligent Transportation System1 Introduction 1
1.1 Motivation 1
1.2 Related Work 2
1.3 Contribution 3
1.4 Thesis Outline 4
2 Vehicle Network Structure and Data Collection 5
2.1 C-ITS 5
2.2 Autonomous Vehicle Data Collection Based on Custom Roads 7
2.3 Autonomous Vehicle Data Collection Based on Actual Roads 9
2.4 Attack Model 10
3 Vehicle Network Attack Detector and Evaluation Method 25
3.1 Attack Detector Structure 25
3.2 Evaluation Method 27
3.3 Neural Network Module 28
3.4 Plausibility Check Module 30
3.5 Attack Detector Design 33
4 Detector Performance Evaluation 37
4.1 Attack Detection Performance using Custom Roads Data 37
4.2 Attack Detection Performance using Actual Roads Data 37
4.3 Attack Detection Performance Against Expanded Attacks 38
5 Conclusion 43
bibliography 47
국문초록 49MasterdCollectio
Practical algorithms for weakly flexible job scheduling for smart mold component process
No full textIn this paper, we focus on a novel troublesome practical challenge, termed Weakly Flexible Job-Shop Scheduling (WFJSP) with parallel machines, where we are required to schedule the jobs with many unconventional limitations, including maximum machine usage constrained by computing resources (single server with limited memory), machines (i.e., idle time, office hours and efficiency), mold components (i.e., idle time, counts, and uncertain processes), and processes (i.e., types and orders) on self-developed intelligent mold processing system hosted by a single resource-limited server. We first highlight the pros and cons of pure theoretical works by comparing with multiple jop-shop scheduling methods, then emphasize the necessity of considering system resource utilization from perspective of software quality. We then shed light on the definitions of different job-shop scheduling problems and clarify the novel scenario at length with six innate conflicts of WFJSP. Based on these detailed analysis, three methods are devised and designed inspired by combining greedy algorithm, ranking strategy, mature infrastructures and well-designed system hierarchy. Experiments are conducted on our self-developed intelligent mold processing system. As a baseline, we introduce a genetic algorithm specifically designed for WFJSP, named WFJSP-GeneA. To evaluate the efficacy and practicability of these four approaches, we adopt five metrics dependent on the software quality attributes. Our proposed algorithms outperform on all the metrics compared with the baseline WFJSP-GeneA. Particularly, we observe outstanding performance of our algorithms on the metrics of Parsimony, Reliability, and Performance. We thus consider that proposed algorithms overcome WFJSP and are practically applicable on production system using mature infrastructures and well-designed system hierarchy.FALSEsciescopu
The single RRM domain-containing protein SARP1 is required for establishment of the separation zone in Arabidopsis
No full textAbscission is the shedding of plant organs in response to developmental and environmental cues. Abscission involves cell separation between two neighboring cell types, residuum cells (RECs) and secession cells (SECs) in the floral abscission zone (AZ) in Arabidopsis thaliana. However, the regulatory mechanisms behind the spatial determination that governs cell separation are largely unknown. The class I KNOTTED-like homeobox (KNOX) transcription factor BREVIPEDICELLUS (BP) negatively regulates AZ cell size and number in Arabidopsis. To identify new players participating in abscission, we performed a genetic screen by activation tagging a weak complementation line of bp-3. We identified the mutant ebp1 (enhancer of BP1) displaying delayed floral organ abscission. The ebp1 mutant showed a concaved surface in SECs and abnormally stacked cells on the top of RECs, in contrast to the precisely separated surface in the wild-type. Molecular and histological analyses revealed that the transcriptional programming during cell differentiation in the AZ is compromised in ebp1. The SECs of ebp1 have acquired REC-like properties, including cuticle formation and superoxide production. We show that SEPARATION AFFECTING RNA-BINDING PROTEIN1 (SARP1) is upregulated in ebp1 and plays a role in the establishment of the cell separation layer during floral organ abscission in Arabidopsis. © 2024 The Author(s). New Phytologist © 2024 New Phytologist Foundation.FALSEsciescopu
Decomposition of Respiratory and Cardiac Rates Using Dual Band Distributed Radar
No full textRespiratory and cardiac rates can be estimated by analyzing a spectrum of phase modulation in a radar echo of an individual. However, the phase modulations corresponding to respiratory and cardiac rates are linearly mixed in a radar echo, so cardiac rate estimation is significantly difficult because of high order harmonics of a respiratory rate. We exploit independent component analysis (ICA) method for effective decomposition of respiratory and cardiac rates with a dual band distributed radar. In simulations and experiments, the respiratory and cardiac rates were successfully estimated by proposed decomposition method, compared with conventional methods. © 2024 European Microwave Association (EuMA)
Enhanced magnetization by defect-assisted exciton recombination in atomically thin CrCl3
No full textTwo-dimensional semiconductors present unique opportunities to intertwine optical and magnetic functionalities and to tune these performances through defects and dopants. Here, we integrate exciton pumping into a quantum sensing protocol on nitrogen-vacancy centers in diamond to image the optically induced transient stray fields in few-layer, antiferromagnetic CrCl3. We discover that exciton recombination enhances the in-plane magnetization of the CrCl3 layers, with a predominant effect in the surface monolayers. Concomitantly, time-resolved photoluminescence measurements reveal that nonradiative exciton recombination intensifies in atomically thin CrCl3 with tightly localized, nearly dipole-forbidden excitons and amplified surface-to-volume ratio. Supported by experiments under controlled surface exposure and density functional theory calculations, we interpret the magnetically enhanced state to result from a defect-assisted Auger recombination that optically activates electron transfer between water vapor related surface impurities and the spin-polarized conduction band. Our work validates defect engineering as a route to enhance intrinsic magnetism in single magnetic layers and opens an experimental platform for studying optically induced, transient magnetism in condensed matter systems. © 2024 American Physical Society.FALSEsciescopu
Enhencement of magnetic properties of Tb-Al-Cu diffusion magnets depending on grain boundary diffusion process conditions
No full textThe sintered Nd-Fe-B magnets have been widely used in the fields of a wide range of applications, such as hard disc drives, magnetic sensor, wind power generators, efficient air-conditioner compressors and motors for electric vehicles. Nd-Fe-B sintered magnets have a large maximum magnetic energy product, but in high-temperature environments such as the motors of eco-friendly cars, the coercivity decreases and performance deteriorates. To achieve high coercivity and remanence at high temperatures, heavy rare-earth (HRE) based grain boundary diffusion process (GBDP) is widely used. In this study, the temperature and time conditions of the GBDP process were optimized using a low-melting point HRE diffusion material based on Nd-Fe-B sintered magnets with Al and Cu added. The magnetic properties and microstructure of Tb-Al-Cu diffusion magnets with various compositions are discussed
JDEC: JPEG Decoding via Enhanced Continuous Cosine Coefficients
No full textWe propose a practical approach to JPEG image decoding, utilizing a local implicit neural representation with continuous cosine formulation. The JPEG algorithm significantly quantizes discrete cosine transform (DCT) spectra to achieve a high compression rate, inevitably resulting in quality degradation while encoding an image. We have designed a continuous cosine spectrum estimator to address the quality degradation issue that restores the distorted spectrum. By leveraging local DCT formulations, our network has the privilege to exploit dequantization and upsampling simultaneously. Our proposed model enables decoding compressed images directly across different quality factors using a single pre-trained model without relying on a conventional JPEG decoder. As a result, our proposed network achieves state-of-the-art performance in flexible color image JPEG artifact removal tasks. Our source code is available at https://github.com/WooKyoungHan/JDEC
Study on Alzheimer's Disease Subtypes Using Generative Adversarial Networks
No full textAlzheimer’s Diseases, Microglia, Generative Adversarial Network simulation, Bioinformatics, ImmunityAlzheimer's disease (AD) is a neurodegenerative disorder characterized by the aggregation of β-amyloid (Aβ)-containing extracellular plaques and tau-containing intracellular neurofibrillary tangles. Various methods are employed to study AD, and in this research, AI technologies, including machine learning and deep learning, were utilized. We identified differentially expressed genes (DEGs) between the 5xFAD mouse 4 months group and 5xFAD mouse 8 months group, revealing a trend toward subtyping within the 5xFAD mouse 8 months group. Using K-means clustering, we divided the 5xFAD mouse 8 months group into two distinct subgroups. Subsequently, Generative Adversarial Networks (GANs) algorithm was employed to simulate the progression from the 5xFAD mouse 4 months group to the two 5xFAD mouse 8 months subgroups. The results indicated that the disease progresses independently into each stage rather than transitioning through one subgroup. We designated these stages as Type 1 and Type 2. Gene Ontology (GO) analysis of significant genes used in the simulation revealed a prevalence of immune-related GO terms. Differences in immune activity exist in human AD subtypes, and for accurate research, it is essential to distinguish subtype models in the 5xFAD mouse. Therefore, this study suggests the potential for more refined and detailed AD research.
Keywords: Alzheimer’s Diseases, Microglia, Generative Adversarial Network simulation, Bioinformatics, Immunity|생성적 적대 신경망을 통한 알츠하이머병의 유형 연구
알츠하이머병은 베타-아밀로이드(Aβ)를 함유한 세포 외 플라크와 타우를 함유한 세포 내 신경원섬유엉킴을 특징으로 하는 신경 퇴행성 질환이다. 알츠하이머병을 연구하기 위해서는 다양한 방법이 사용되어 왔으며, 본 연구에서는 머신 러닝 및 딥러닝을 포함한 인공지능 기술을 사용하여 연구를 수행하였다.5xFAD 마우스 4개월 그룹과 8개월 그룹 간의 차등 발현 유전자 (DEGs) 를 확인하였고, 5xFAD 마우스 8 개월 그룹 내에 하위 그룹이 관찰되는 경향성을 밝혀냈다. K-means 클러스터링을 사용하여 5xFAD 마우스 8 개월 그룹을 두 개의 하위 그룹으로 나누었다. 이를 기반으로 생성적 대립 신경망 (GANs) 알고리즘을 사용하여 5xFAD 4 개월 그룹에서 5xFAD 8 개월 그룹 1 및 그룹 2 로의 진행을 모사했다. 시뮬레이션 결과, 4 개월 그룹에서 8 개월 그룹으로 도착할 때 특정 하위 그룹을 거쳐서 진행되는 것이 아니라, 각각의 단계로 독립적으로 진행되는 것을 확인하였다. 우리는 이 8 개월 그룹의 두 단계를 Type 1 과 Type 2 로 명명하였다. 이어서, 시뮬레이션에서 사용된 주요 유전자에 대한 유전자 온톨로지 (GO) 분석을 수행하였다. 다양한 GO 항목 중 면역과 관련된 항목이 풍부하게 나타난 것을 확인하였다. 우리는 Type 1 과 Type 2 를 나누는 데 중요한 역할을 하는 유전자들을 확인하였고, 이 유전자들은 생체의 면역 활동과 밀접하게 관련된 것으로 보인다. 인간의 AD 하위 그룹에서도 면역 활동에 의한 차이가 존재하며, 정확도 높은 연구를 위해 5xFAD 마우스에서 하위 그룹 모델을 구별하여 연구해야 한다. 따라서 본 연구는 더 정교하고 섬세한 AD 연구의 가능성을 시사한다.
핵심어: 알츠하이머, 미세아교세포, 생성적 대립 신경망, 생명정보학, 면역Ⅰ. Introduction 1
1.1 Microglia in AD research 1
1.2 AD subtypes 2
1.3 Analysis techniques 3
1.3.1 Next-generation sequencing technology 3
1.3.2 Generative Adversarial Networks 4
II. Methods and Materials 6
2.1 RNA sequencing data processing 6
2.2 K-means clustering 6
2.3 Generative Adversarial Network simulation 7
2.3.1 GANs 7
2.3.2 Data preprocessing 7
2.3.3 GANs algorithm 8
2.4 Latent interpolation and latent extrapolation 8
2.4.1 Latent interpolation 8
2.4.2 Latent extrapolation 9
2.5 Gene selection 9
2.6 Functional enrichment analysis 10
III. Results 11
3.1 Identification of Tendency for Division into Two Subgroups within the 5xFAD 8 Months Group 13
3.2 Utilizing DEGs for Preparing GANs Simulation 16
3.3 Conducting GAN Simulation from 5xFAD Mouse 4 Months Group to 8 Months Group 19
3.4 Functional Enrichment Analysis with Intersection Genes 22
IV. Discussion 34
V. References 37
Abstract in Korean 40MasterdCollectio