GIST Scholar
Not a member yet
30271 research outputs found
Sort by
제조업 현장에서 딥러닝 프로젝트 실무 워크플로우 제안
No full text제조업에서 인공지능(AI) 기술 도입은 생산 효율 제고, 불량률 감소, 예방 정비 구현을 목표 로 빠르게 확산되고 있다. 그러나 제조 공정의 복잡성, 이해관계자의 다양성, 데이터 보안 제약 등으로 인해 딥러닝 프로젝트는 예산 초과, 일정 지연, 노사 갈등에 직면하기 쉽다. 본 연구는 이 러한 위험 요인을 최소화하기 위해 (1) 전략적 투자 기획, (2) 신속 PoC, (3) 데이터·모델 개발, (4) 배포·운영, (5) 사후 성과 분석의 다섯 단계로 구성된 실무 지향 워크플로우를 제안하고, 각 단계 에 RACI(Responsible, Accountable, Consulted, Informed) 매트릭스를 적용하여 권한과 책임을 명확히 규정하였다. 특히, 전략적 투자 기획 단계에서는 이해관계자의 요구사항과 예산 제한사항 을 반영하여 현실성 높은 프로젝트 목표를 설정하고, 신속 PoC 단계에서는 초기 타당성 검증을 통해 기술적 위험을 사전에 최소화하였다. 데이터·모델 개발 단계에서는 현장의 실제 데이터를 수 집하고 체계적인 품질관리를 수행하여 모델의 정확성과 신뢰성을 높이는 방법을 제시한다. 배포· 운영 단계에서는 현장 상황에 맞는 인프라를 선택하고, CI/CD 기반 자동화 시스템을 구축하여 안 정성과 효율성을 강화하였다. 사후 성과 분석 단계에서는 지속적인 모니터링과 자동 재학습 루프 를 통해 운영 중 발생 가능한 모델 성능 저하에 적극적으로 대응하였다. 워크플로우의 실효성은 금속 표면 결함 검출 사례(YOLOv8 기반)를 통해 검증하였다. 본 연구는 노사 협상, 인력 재배치, 산업 기밀 보호 등 현장 현실을 System 내에 내재화함으로써 기존 기술 중심 가이드라인의 한계 를 보완하였다. 워크플로우는 공정 복잡도와 기업 규모에 상관없이 적용 가능하며, 제조업의 AI 전환을 가속화하고 프로젝트 성공률을 제고할 수 있는 재현 가능한 청사진을 제공한다. 구체적인 워크플로우 내용은 3장에서 각 단계(3.1.1, 3.2.1 등)의 세부 내용을 통해 확인할 수 있다.|The adoption of artificial intelligence (AI) technology in the manufacturing industry is rapidly increasing to enhance production efficiency, reduce defect rates, and implement predictive maintenance. However, due to the complexity of manufacturing processes, diversity of stakeholders, and data security constraints, deep learning projects often face risks such as budget overruns, schedule delays, and labor-management conflicts. To minimize these risks, this study proposes a practitioner-oriented workflow composed of five phases: (1) Strategic Investment Planning, (2) Rapid Proof of Concept (PoC), (3) Data and Model Development, (4) Deployment and Operations, and (5) Post-Deployment Performance Analysis. Each phase clearly defines roles and responsibilities using the RACI (Responsible, Accountable, Consulted, Informed) matrix.
Specifically, in the Strategic Investment Planning phase, realistic project objectives are set by considering stakeholder requirements and budget limitations. The Rapid PoC phase mitigates technical risks through early feasibility verification. The Data and Model Development phase outlines methods for collecting real-world data from the field and performing systematic quality control to enhance model accuracy and reliability. In the Deployment and Operations phase, infrastructure tailored to on-site conditions is selected, and stability and efficiency are improved by establishing a CI/CD-based automated system. The Post-Deployment Performance Analysis phase actively responds to potential model
performance degradation through continuous monitoring and automated retraining loops. The practical effectiveness of the proposed workflow is validated through a case study involving the detection of metal surface defects using the YOLOv8 model. This study integrates workplace realities such as labor negotiations, workforce reallocation, and protection of industrial secrets into the workflow, addressing limitations of traditional technology-centered guidelines. The proposed workflow is applicable regardless of process complexity or company size, offering a reproducible blueprint that can accelerate AI transformation in manufacturing and improve project success rates. Detailed workflow descriptions for each phase (e.g., sections 3.1.1, 3.2.1, etc.) can be found in Chapter 3.MasterI. 서론 1
1.1 연구배경 및 목적 1
1.2 관련연구 2
1.2.1 AI 프로젝트에서의 이해관계자 역할 2
1.2.2 RACI 기반 이해관계자 역할 및 책임 정의 연구 3
1.2.3 제조업에서의 AI 프로젝트 사례 3
1.3 논문 범위 및 구성 4
II. 산업현장에서 고려해야할 사항 5
2.1 Business 측면 5
2.1.1 이해관계자 선정 및 업무 분담 5
2.1.2 딥러닝 적용을 통한 해결 목표 설정 8
2.1.3 ROI (Return On Investment) 분석 9
2.2 딥러닝 기술 요소 10
2.2.1 데이터 11
2.2.2 모델 11
2.2.3 배포 및 운영 인프라 구축 13
2.3 조직 및 인력 관리 17
III. 제조업 딥러닝 프로젝트를 위한 워크플로우 설계 18
3.1 프로젝트 기획 및 투자 18
3.1.1 프로젝트 목표 및 범위 설정 19
3.1.2 이해관계자 선정 및 조직 구성 19
3.1.3 투자 기획 및 예산 편성 20
3.1.4 투자 일정 수립 20
3.1.5 인력변동 파악 및 노사 협의 21
3.1.6 투자의결 및 리스크 관리 21
3.2 AI 모델 구축 및 운영 21
3.2.1 [Step1 Quick PoC] 공개 데이터셋 및 모델 검토 22
3.2.2 [Step1 Quick PoC] 모델 학습 및 평가 23
3.2.3 [Step1 Quick PoC] 결과 분석 23
3.2.4 [Step1 Quick PoC] 초기 타당성 검토 23
3.2.5 [Step2 데이터&모델 구축] 데이터 거버넌스 및 보안 25
3.2.6 [Step2 데이터&모델 구축] 공정용 실데이터 취득 25
3.2.7 [Step2 데이터&모델 구축] 품질관리 및 라벨링 25
3.2.8 [Step2 데이터&모델 구축] 데이터 관리 및 리스크 검토 26
3.2.9 [Step2 데이터&모델 구축] 모델 선정 및 개발 26
3.2.10 [Step2 데이터&모델 구축] 모델 학습 및 평가 26
3.2.11 [Step2 데이터&모델 구축] 파일럿 적용 및 결과 분석 27
3.2.12 [Step3 운영체계 수립 및 배포] 배포 전략 수립 27
3.2.13 [Step3 운영체계 수립 및 배포] 지속통합(CI) 테스트 28
3.2.14 [Step3 운영체계 수립 및 배포] MLOps 파이프라인 구축 29
3.2.15 [Step3 운영체계 수립 및 배포] 운영리스크 관리 및 결과분석 30
3.3 사후관리 및 성과분석 30
3.3.1 성과지표 모니터링 및 재무성과 분석 31
3.3.2 현업 이관 및 교육 31
3.3.3 새로운 공정으로의 확대적용 32
IV. 적용 사례: 표면검사 프로젝트 33
4.1 프로젝트 목표설정 33
4.2 Quick Proof of Concept (PoC) 34
4.2.1 Experiments 34
4.2.2 초기 타당성 검토 38
4.3 프로젝트 기획 및 투자 38
4.4 데이터 및 모델 구축 40
4.5 운영 및 배포 40
4.6 사후관리 및 성과분석 41
V. 결론 42
참고문헌 4
Electrolyte additive design to address side reactions at cathode and anode in flowless aqueous Zn-Br2 batteries
No full textAqueous Zn-Br2 batteries hold immense promise for large-scale energy storage systems due to their inherent safety and high energy density. However, achieving high battery performance necessitates addressing critical side reactions, including dendrite formation and hydrogen evolution at the anode, as well as bromine species diffusion and evaporation of volatile bromine at the cathode. Importantly, these reactions impact overall battery performance, underscoring the need for comprehensive mitigation strategies. This study proposes an electrolyte additive design strategy to address side reactions at both the anode and cathode. This study introduces a multi-type electrolyte additive strategy combining cerium chloride (CeCl3) and 1-ethylpyridinium bromide (1-EPBr). Trivalent Ce3+ forms an electrostatic shielding layer to prevent Zn2+ from concentrating at zinc metal protrusions, while the high electron-donating nature of Cl− mitigates H2O decomposition on the zinc metal surface by reducing the interaction between Zn2+ and H2O. These combined cationic and anionic effects significantly enhance the reversibility of the zinc metal reaction. Additionally, 1-EPBr, a known bromine complexing agent (BCA), effectively mitigates bromine species diffusion and the evaporation of bromine at the cathode. This enables the flowless aqueous Zn-Br2 battery to reliably cycle with exceptionally high capacity (>400 mAh after 5,000 cycles) even in a large-scale battery configuration of 15 × 15 cm2. This work presents a novel approach by utilizing both cationic and anionic additives to stabilize the zinc metal anode. These findings advance the understanding of stabilization mechanisms in aqueous battery systems. Building upon this foundation, this study proposes a single-type electrolyte additive to simultaneously address critical challenges related to both anode and cathode reactions in aqueous Zn-Br2 batteries. This work introduces poly(N-(2-hydroxyethyl)-4-vinylpyridinium bromide), a multifunctional poly(ionic liquid) (PIL). The pyridinium moieties within the PIL, featuring electron-deficient aromatic rings, establish π-anion interactions with polybromide anions (Br3 −, Br5 −, Br7 −, etc.), effectively mitigating bromide diffusion and evaporation. Additionally, the PIL electrostatically adsorbs onto the Zn anode, forming an electro-shielding layer that is expected to inhibit Zn dendrite formation. The hydroxyl (−OH) functional groups within the PIL further enhance its functionality by participating in Zn2+ solvation structures, thereby reducing the HER at the Zn anode through modulation of solvation dynamics. This marks the first application of a PIL as an electrolyte additive in aqueous Zn-Br2 batteries. This study highlights the potential of PIL-based additives for extending to other aqueous battery systems, paving the way for next-generation, high-performance energy storage technologies.MasterAbstract i
Contents ii
List of figures iii
1. Introduction 1
2. Experimental 3
2.1. CeCl3 additive 3
2.1.1. Preparation of electrolytes 3
2.1.2. Characterization 3
2.1.3. Battery assembly 3
2.1.4. Electrochemical measurements 4
2.2. Poly(N-(2-hydroxyethyl)-4-vinylpyridinium bromide) additive 4
2.2.1. Preparation of N-(2-hydroxyethyl)-4-vinylpyridinium bromide 4
2.2.2. Preparation of poly(N-(2-hydroxyethyl)-4-vinylpyridinium bromide) 4
2.2.3. Characterization 5
2.2.4. Preparation of electrolytes 5
2.2.5. Battery assembly 5
2.2.6. Electrochemical measurements 5
3. Results and discussion 6
3.1. CeCl3 additive 6
3.1.1. Electrochemical test of a graphite (cathode)/Zn (anode) cell 6
3.1.2. Analysis of zinc surface and bulk electrolyte 6
3.1.3. Electrochemical test of a graphite (cathode)/ graphite (anode) cell 8
3.2. Poly(N-(2-hydroxyethyl)-4-vinylpyridinium bromide) additive 24
3.2.1. Synthesis and characterization 24
3.2.2. Properties of poly(N-(2-hydroxyethyl)-4-vinylpyridinium bromide) 24
3.2.3. Influence of the synthesized PIL on the cathode 25
3.2.4. Influence of the synthesized PIL on the anode 25
3.2.5. Electrochemical test of a graphite (cathode)/ graphite (anode) cell 26
4. Conclusion 37
5. References 38
6. 감사의 글 4
ConnecToMind: Connectome-Aware fMRI Decoding forVisual Image Reconstruction
No full textRecent deep-learning approaches have achieved significant improvements in reconstructing visual images from human brain activity. However, existing methods typically represent brain activity as flattened voxel-wise signals, overlooking the detailed anatomical and functional organization of visual cortical regions. Here, we propose ConnecToMind, a novel decoding framework that employs a region-level fMRI embedding module to preserve distinct functional representations across visual cortical sub-regions, while leveraging functional connectivity (FC) derived from resting-state fMRI. Experiments on the Natural Scenes Dataset (NSD) demonstrate that ConnecToMind outperforms the MindEye in both the semantic and perceptual fidelity of reconstructed images, validating the effectiveness of preserving distinct functional representations with FC prior. Moreover, ConnecToMind shows competitive performance in image retrieval tasks. Ablation analyses further reveal that low-level (e.g., V1–V3) and high-level (e.g., Lateral Occipital, Fusiform) visual regions distinctly contribute to the reconstruction quality, highlighting the importance of region-specific embeddings in visual reconstruction. All codes for this study are publicly available at GitHub (https://github.com/aimed-gist/ConneToMind)
Potential influence of ammonia reduction on particulate nitrate concentration in South Korea
No full textIn Northeast Asia, nitrate concentration in fine particles has little decreased even with a significant reduction in nitrogen oxides in recent years. Ammonia reduction can be considered as one of the control strategies for nitrate. This study investigated the role of ammonia in the formation of nitrate in the ambient atmosphere by using various observation data at 10 locations (urban-agricultural, urban-industrial, urban, and remote) in South Korea in 2022 and the E-AIM thermodynamic equilibrium model. In the warm season (May-October), ammonia had a significant effect on particulate nitrate concentration by nitric acid-nitrate partitioning. A sensitivity test suggested that on average across all locations in the warm season, a 50% reduction in ammonia can lead to a 47% decrease in particulate nitrate with a constant total nitrate concentration (= gaseous nitric acid + particulate nitrate). This effect was the weakest at the agricultural site with high ammonia level (19.0 ppb). In the cold season (November-April), the ammonia reduction was not effective in controlling the nitrate by the partitioning, and long-range transport affected nitrate concentration at the remote site. With the consideration of the rapid loss of the gas-phase species in the total nitrate, the partitioning with ammonia reduction was also helpful to control the nitrate. Our findings suggest that ammonia emission reduction via nitric acid-nitrate partitioning should be an effective strategy to decrease particulate nitrate level in the ambient atmosphere particularly in the warm season. © 2025 Elsevier B.V.FALSEsciescopu
Modifying electronic and structural properties of 2D van der Waals materials via cavity quantum vacuum fluctuations: a first-principles QEDFT study [Invited]
No full textStructuring the photon density of states and light-matter coupling in optical cavities has emerged as a promising approach to modifying the equilibrium properties of materials through strong light-matter interactions. In this article, we employ state-of-the-art quantum electrodynamical density functional theory (QEDFT) to study the modifications of the electronic and structural properties of two-dimensional (2D) van der Waals (vdW) layered materials by the cavity vacuum field fluctuations. We find that cavity photons modify the electronic density through localization along the photon polarization directions, a universal effect observed for all the 2D materials studied here. This modification of the electronic structure tunes the material properties, such as the shifting of energy valleys in monolayer h-BN and 2H-MoS2, enabling tunable band gaps. Also, it tunes the interlayer spacing in bilayer 2H-MoS2 and Td-MoTe2, allowing for adjustable ferroelectric, nonlinear Hall effect, and optical properties, as a function of light-matter coupling strength. Our findings open an avenue for engineering a broad range of 2D layered quantum materials by tuning vdW interactions through fluctuating cavity photon fields.TRUEsciescopu
State-wise Safety in Autonomous Driving via Lagrangian-based Constrained Reinforcement Learning
No full textFor the practical deployment of autonomous driving systems, high levels of safety and adaptability are essential. Accordingly, Deep Reinforcement Learning (DRL), which learns and improves driving strategies through trial and error, has gained attention. However, the reward-driven nature of reinforcement learning may still lead to unsafe or abnormal behavior even after training. To address this limitation, Constrained Reinforcement Learning (CRL) has been proposed to balance safety and performance. While CRL typically defines constraints as expected cumulative costs, this formulation does not consider whether constraints are satisfied at each state, making it difficult to ensure state-wise safety. In this paper, we extend a Lagrangian-based CRL approach by estimating state-wise Lagrangian multipliers, allowing the policy to account for state-level safety. We evaluate the proposed method in OpenAI's Safety Gym environment and compare its performance with existing Lagrangian-based methods.|자율주행 시스템의 실제 적용을 위해서는 높은 안정성과 적응성이 요구된다. 이에 따라, 시행착오를 통해 주행 전략을 학습하며 발전시키는 심층 강화 학습(Deep Reinforcement Learning, DRL)이 주목받고 있다. 하지만 강화 학습은 본질적으로 보상을 극대화하는 방향으로 정책을 학습하기 때문에, 학습 후에도 안전하지 않거나 비정상적인 행동을 할 가능성을 완전히 배제하기 어렵다. 이러한 한계를 해결하기 위해, 정책 학습 시 안정성과 성능 간의 균형을 도모하는 제약 강화 학습(Constrained Reinforcement Learning, CRL)이 제안되었다. 제약 강화 학습은 기댓값 기반 누적 비용 형태의 제약 조건을 만족하도록 정책을 학습하지만, 각 상태에서의 제약 조건 충족 여부를 고려하지 않아 상태별 안정성을 보장하기 어렵다. 본 논문에서는 제약 강화 학습의 한 방식인 라그랑지안 기반의 방법을 확장하여, 상태별 라그랑주 승수를 추정함으로써 정책이 상태별 안정성을 고려하도록 한다. 또한 제안한 방법을 OpenAI의 시뮬레이션 환경인 Safety Gym을 통해 기존 라그랑지안 기반의 방법들과 비교하여 검증하였다.Master1. Introduction 1
1.1 Introduction 1
1.2 Research Objective 3
1.3 Outline of the Thesis 4
2. Background 5
2.1 Reinforcement Learning 5
2.1.1 Policy Gradient Methods 6
2.1.2 Off-Policy Gradient Methods 11
2.2 Constrained Reinforcement Learning 12
2.2.1 Lagrangian Method 13
2.2.2 Related Work: PPO Lagrangian 14
2.3 State-wise Constrained Reinforcement Learning 15
2.3.1 Related Work: Feasible Actor-Critic 16
3. PPO-based Method in State-wise Constrained RL 19
3.1 PPO Lagrangian Network 19
3.1.1 Comparison with PPO Lagrangian 21
3.1.2 Comparison with Feasible Actor-Critic 21
4. Experiments 23
4.1 Setup 23
4.2 Analyzing the Influence of alpha in the Feasible Actor-Critic 25
4.3 Analyzing the Influence of Bias Initialization in the Lagrange Multiplier Network of PPO Lagrangian Network 28
4.4 Analyzing the Influence of Learning Rate in the Lagrange Multiplier Network of PPO Lagrangian Network 31
4.5 Evaluation Results 34
4.5.1 Car Goal 34
4.5.2 Car Button 37
4.6 Evaluation Lagrange Multiplier Network 40
5. Conclusion 43
5.1 Conclusion 43
5.2 Limitations and Future Work 44
Summary 45
References 46
Acknowledgements 5
A novel scheme for speed variation of a robotic cane to improve step length symmetry during overground walking
No full textGait symmetry is a vital index for improving and evaluating the walking ability of patients with hemiplegia. Asymmetric gait disturbs walking balance and increases unnecessary energy consumption, causing serious damage to the musculoskeletal system. The restoration of step length symmetry (SLS) is one of the gait parameters targeted for rehabilitation and can be easily accessed and adjusted in the clinic. Therefore, we proposed a novel scheme based on a speed variation method (SVM) using a robotic haptic cane (HC) to improve SLS on overground walking. To evaluate the proposed method and check protocols for a patient study, we conducted an experiment to lengthen the left step length of 24 healthy young participants using positive (P), negative (N), and positive & negative (PN) SVM. Additionally, we examined its feasibility in one stroke patient. As a result of the healthy subjects' experiment, walking speed and SLS ratio significantly increased with P and PN SVM while maintaining Root Mean Square (RMS) of pelvic tilt (p < .001). Results from the patient pilot test were similar to those of the healthy subjects, with decreased RMS of pelvic tilt. SLS ratio improved from-0.13 +/- 0.00 (Normal Walking) to-0.07 +/- 0.01 (P SVM) and-0.04 +/- 0.01 (PN SVM). However, swing phase symmetry also improved. Furthermore, the patient was sensitive to the baseline HC speed and Negative SVM. The results indicate that the proposed methods may be effective in stroke patients and the relationship between walking speed and the ratio of PN SVM should be verified in a larger number of patients.TRUEsciescopu
Practical approaches to apply Human Pose Estimation for real-world applications
No full textHuman Pose Estimation (HPE) has emerged as a vital area of research in computer vision, aiming to estimate human body configurations from input visual data. This dissertation addresses the challenges of HPE by proposing a set of practical methodologies focused on enhancing efficiency, robustness, and real-world applicability. The research is structured around three interconnected studies, each tackling a unique aspect of pose estimation, yet collectively contributing to the broader objective of developing scalable and high-performing HPE systems. The first study centers on multi-view 3D HPE, introducing a lightweight and markerless skeleton tracking algorithm that effectively resolves self-occlusion—a persistent challenge in pose estimation. This is achieved by merging pose candidates derived from multiple RGB-D sensors using a combination of DBSCAN clustering and Kalman filtering. By avoiding reliance on heavy deep learning models, the proposed algorithm is suitable for real- time applications in resource-constrained environments. Experimental evaluations confirm its superiority in tracking limb joints under occlusion, underscoring the importance of sensor fusion and spatial redundancy. However, this approach still requires a careful sensor installation process and may suffer from reduced accuracy due to the inherent limitations of depth information, such as susceptibility to environmental interference (e.g., sunlight or reflective surfaces), as well as the challenges posed by suboptimal sensor placement or simultaneous tracking of multiple individuals. Building upon the foundational insights from the first study, the second study transitions from controlled sensor environments to consumer-grade RGB videos, applying 3D human reconstruction (HR) techniques in the context of dance education. The resulting system, DanceSculpt (DS), demonstrates how HR can deliver multi-angle visualizations of human movement without the limitations of depth-based systems, such as IR interference and complex calibration. By leveraging 3D avatars reconstructed from monocular input, DS provides learners with detailed visual feedback, improving understanding of posture, timing, and spatial formation. The successful application of this HR method in dance learning emphasizes its potential for other motion-intensive educational contexts, creating a direct link to the goals of the first study while expanding its practical relevance. Nonetheless, DS’s reliance on a top-down approach results in increased inference time proportional to the number of detected individuals, making real-time performance a challenge. Furthermore, its inability to reconstruct poses accurately under severe occlusion or when most body parts are not visible remains a significant limitation. The third study builds upon the system-level insights from the first two investigations by introducing an advanced yet efficient 2D multi-person pose estimation framework that enhances performance within a one-stage architecture. Rather than generalizing the findings, this study focuses on achieving additional performance gains through a more effective integration of instance-centric attention mechanisms. InstaPose, developed in this stage, incorporates a novel Instance-Centric Keypoint Attention (ICKA) mechanism within a DETR-based transformer model. This design directly addresses a key limitation observed in earlier approaches—insufficient interaction between instance and keypoint queries—by enhancing contextual coherence and spatial precision. Extensive experiments on MS COCO and CrowdPose datasets validate the framework’s effectiveness, demonstrating its superiority in crowded scenes with minimal parameter overhead. The performance gains achieved here reflect lessons learned from both the robust merging strategies of the first study and the reconstruction-based feedback system of the second. However, InstaPose still inherits limitations of transformer-based architectures, including relatively high computational cost and complexity in training. Moreover, the framework lacks extension to 3D HPE, which limits its application in scenarios requiring full spatial understanding. Together, these three studies form a cohesive research trajectory that progressively abstracts from multi-sensor integration to high-level model architecture. This studies highlights how HPE solutions can be adapted across varying input modalities and use cases, from high-precision tracking systems to educational and real-time applications. This dissertation contributes to the ongoing evolution of HPE by showing that accurate, efficient, and scalable solutions are not mutually exclusive but can be simultaneously realized through thoughtful system design and cross-domain insight. The proposed approaches hold promise for a wide range of applications, including interactive learning, health monitoring, sports analysis, and beyond, where understanding and interpreting human motion is essential.DoctorChapter 1 １
INTRODUCTION １
1.1. Background １
1.2. Thesis Overview ２
Chapter 2. ４
Accurate 3D pose tracking by merging multiple pose from multi-view sensors ４
2.1. Introduction : 3D Human Pose Estimation ５
2.3. Arrangement of Skeleton to Correct the Misoriented Joints １０
2.4. Skeleton Merging and Noise Filtering Using DBSCAN １１
2.5. Experimental settings １３
2.6. Results １６
2.7. Discussion ２０
2.8. Conclusion and Limitations ２４
Chapter 3 ２６
Human reconstruction based assistance tools for dance learning ２６
3.1 Introduction : Dance assistant tools for learners ２６
3.2 Related works : Dance learning tools and Human reconstruction ２８
3.3 DanceSculpt ３２
3.4 Experimental Design ３７
3.6 Conclusion and Limitations ４４
Chapter 4 ４６
Efficient 1-stage 2D pose estimation based DETR architecture ４６
4.1. Introduction : 2D Human pose estimation ４７
4.2. Related works : One-stage 2D Human pose estimation ４９
4.3. Preliminary and Overview of InstaPose ５２
4.4. Instance-Centric keypoint Attention (ICKA) ５５
4.5. Experimental Setup ５７
4.6. Result ５９
4.7. Ablation Study ６１
4.8 Discussion ６２
4.9 Conclusion ６３
Chapter 5 ６５
Discussion ６５
Chapter 6 ７０
Conclusion ７０
References ７１
Appendix A. ７７
1. Calibration for Coordinate Systems of Sensors ７７
1.1. Sensor-to-Sensor Calibration ７８
1.2. Sensor-to-Marker Calibration ７９
2. Joint Position Tracking Using Kalman Filter ８０
3. Comparison with other merging methods ８１
References ８３
Appendix B ８４
References ８５
Appendix C. ８６
1. Comparisons on OCHuman ８６
2. Comparisons on HumanArt ８６
3. Comparisons with Two-stage Methods ８７
4. Further analysis for ICKA ８８
5. Query update types ８９
6. Comparison for efficient with various approaches ９０
7. Comparison for efficiency of proposed algorithm with various 2D HPE ９２
References ９
Integrated DNN-Based Parameter Estimation for Multichannel Speech Enhancement
No full textOne of the popular configurations for the statistical model-based multichannel speech enhancement (SE) is to apply a spatial filter such as the minimum-variance distortionless response beamformer followed by a single channel post-filter, and some of the deep neural network (DNN)-based approaches mimic it. While a number of DNN-based SE focused on direct estimation of clean speech features or the masks to estimate clean speech, some of the efforts were devoted to estimate the statistical parameters. DNN-based parameter estimation with two DNNs for a beamforming stage and a post-filtering stage has demonstrated impressive performance, but the parameter estimation for a beamformer and that for a post-filter operate separately, which may not be optimal in that the post-filter cannot utilize spatial information from multi-microphone signals. In this letter, we propose integrated DNN-based parameter estimation for multichannel SE based on both the beamformer output and multi-microphone signals. The speech presence probability and the power spectral densities for speech and noise estimated in the beamforming stage are utilized in the post-filtering stage for better parameter estimation. We also adopt the dual-path conformer structure with an encoder and decoders to enhance the performance. Experimental results show that the proposed method marked the best wideband perceptual evaluation of speech quality (PESQ) scores on the CHiME-4 dataset among all methods with comparable computational complexity. © 2025 Elsevier B.V., All rights reserved.FALSEsciescopu
Development of Biaryl N-Heterocyclic Carbene Ligands Featuring Non-Covalent Interactions
No full textThe field of organotransition-metal chemistry has extensively studied the development of novel and efficient catalysts for homogeneous catalysis. N-Heterocyclic carbenes (NHCs) are a powerful class of ligands in transition metal catalysis due to their strong electron-donating ability and excellent synthetic flexibility. They have been recognized as such since the stable free carbene was isolated by Arduengo in 1991. Various structural modifications have been employed to adjust the electronic and steric characteristics of NHC ligands, based on the prototypical imidazolylidene skeleton. Imidazo[1,5- a]pyridin-3-ylidene (ImPy) ligands have a rigid bicyclic structure comprising a C5-positioned aryl group. Aryl substituents of the bicyclic ImPy are positioned near a metal coordination sphere, thus allowing for bonding interactions of the metal with substituent. Additionally, ImPy ligands can be synthesized via a concise synthetic route, enabling the convenient incorporation of numerous functional substituents. In line with our interest to developing advanced NHC-based catalysts, we suggest a variety of ligands that feature non-covalent interactions. The present thesis describes the development of imidazopyridine-based NHC transition metal complexes and their application to chemical reactions: diastereoselective addition and desymmetric C-N bond formation. In catalyst design, the Cu-F interaction play a crucial role in stabilizing catalyst and reaction intermediate, thereby influencing the reactivity of catalytic processes. Additionally, fluorine atom, which is not impose spatial demands, can fundamentally alter the performance of the catalyst. The fluorinated aryl groups on the ligand structures also improve the π-back-donation from the metal. Chapter 2 describes the synthesis of fluorinated ImPy ligand (F-ImPy), characterized by a Cu-F interaction and a C5-aryl substituent and their application in the diastereoselective additions of 1,3- PhD/CH 20182101DoctorThe field of organotransition-metal chemistry has extensively studied the development of novel and efficient catalysts for homogeneous catalysis. N-Heterocyclic carbenes (NHCs) are a powerful class of ligands in transition metal catalysis due to their strong electron-donating ability and excellent synthetic flexibility. They have been recognized as such since the stable free carbene was isolated by Arduengo in 1991. Various structural modifications have been employed to adjust the electronic and steric characteristics of NHC ligands, based on the prototypical imidazolylidene skeleton. Imidazo[1,5-a]pyridin-3-ylidene (ImPy) ligands have a rigid bicyclic structure comprising a C5-positioned aryl group. Aryl substituents of the bicyclic ImPy are positioned near a metal coordination sphere, thus allowing for bonding interactions of the metal with substituent. Additionally, ImPy ligands can be synthesized via a concise synthetic route, enabling the convenient incorporation of numerous functional substituents. In line with our interest to developing advanced NHC-based catalysts, we suggest a variety of ligands that feature non-covalent interactions. The present thesis describes the development of imidazopyridine-based NHC transition metal complexes and their application to chemical reactions: diastereoselective addition and desymmetric C-N bond formation.
In catalyst design, the Cu-F interaction play a crucial role in stabilizing catalyst and reaction intermediate, thereby influencing the reactivity of catalytic processes. Additionally, fluorine atom, which is not impose spatial demands, can fundamentally alter the performance of the catalyst. The fluorinated aryl groups on the ligand structures also improve the π-back-donation from the metal. Chapter 2 describes the synthesis of fluorinated ImPy ligand (F-ImPy), characterized by a Cu-F interaction and a C5-aryl substituent and their application in the diastereoselective additions of 1,3-enyne nucleophile to ketones. X-ray and non-covalent interaction (NCI) analysis elucidated both the Cu-arene interaction and the Cu-F interaction in the catalyst. These biaryl ImPy−Cu catalysts were effectively employed in the Cu-catalyzed diastereoselective addition reaction of easily accessible olefins to ketones. F-ImPy-Cu catalyst exhibits higher reactivity than non-fluorinated ImPy-Cu catalysts. Consequently, we achieved tertiary alcohol synthesis with up to 96% yield and up to >10:1 diastereoselectivity.
Anagostic interactions play a crucial role in stabilizing specific ligand and intermediate conformations, thereby influencing the reactivity and selectivity of catalytic processes. In chiral catalyst design, anagostic C-H···M interactions are particularly intriguing as they can restrict the rotation of substituents around the metal center. This restriction is crucial for forming a well-defined chiral environment, essential for achieving high enantioselectivity in catalytic reactions. Notably, this study represents the first attempt to utilize anagostic C-H···M interactions to control enantioselectivity, underscoring its innovative approach. Chapter 3 describes novel chiral biaryl ImPy ligands, characterized by an anagostic C-H···Pd interaction and a C5-aryl substituent and their application as steering ligands in Pd-catalyzed desymmetric C-N bond formation of 3,4-dihydroquinolin-2-ones with quaternary stereocenters. X-ray and NCI analysis elucidated both the Pd-arene interaction and the anagostic interaction in the chiral catalyst. These biaryl ImPy−Pd catalysts were effectively employed in the Pd-catalyzed asymmetric desymmetric C-N cross-coupling of malonamide derivatives. Consequently, we achieved the synthesis of chiral 3,4-dihydroquinolin-2-one derivatives with excellent yields (up to 99%) and high enantioselectivities (up to 97:3 er).
TABLE OF CONTENTS
ABSTRACT···················································································································i
TABLE OF CONTENTS·······························································································iii
LIST OF FIGURES·····································································································vii
LIST OF SCHEMES·····································································································ix
LIST OF TABLES········································································································xi
CHAPTER 1. A Thesis Overview····················································································1
1.1 Non-covalent Interactions in Metal Complex Catalysis··················································2
1.1.1 Intermolecular Non-Covalent Interactions···························································4
1.1.2 Intramolecular Non-Covalent Interactions···························································6
1.2 N-Heterocyclic Carbenes (NHCs)···············································································7
1.3 Imidazo[1,5-a]pyridin-3-ylidene···············································································11
1.3.1 Achiral Imidazo[1,5-a]pyridin-3-ylidene Ligands···············································12
1.3.2 Chiral Imidazo[1,5-a]pyridin-3-ylidene Ligands·················································14
1.3.2.1 Chiral Monodentate Imidazo[1,5-a]pyridin-3-ylidene Ligands·····················15
1.3.2.2 Chiral bidentate Imidazo[1,5-a]pyridin-3-ylidene Ligands···························22
1.4 Thesis Research······································································································27
1.5 References··············································································································28
CHAPTER 2. Fluorinated Biaryl N-Heterocyclic Carbene Ligand with Non-Covalent Interaction for Cu-catalyzed Diastereoselective Addition Reaction································39
2.1 Introduction·············································································································40
2.2 Results and Discussion······························································································42
2.2.1 Synthesis of Ligands and Catalysts····································································42
2.2.2 X-ray Crystallography······················································································43
2.2.3 Non-Covalent Interaction Plot Analysis······························································44
2.2.4 Electronic Properties for F-ImPy Ligands by IR and 77Se-NMR····························45
2.2.5 Catalytic Properties of ImPy-Cu Complexes·······················································47
2.3 Conclusion··············································································································52
2.4 Experimental Section·······························································································53
2.4.1 General Information·························································································53
2.4.2 General Procedure for Synthesis of ImPy Ligands and Catalysts····························54
2.4.3 General Procedures for Diastereoselective Addition with 1,3-Enyne and Ketones···60
2.4.4 X-ray Analysis for Complex 4e·········································································63
2.4.5 Computational Study and Non-Covalent Interaction Plot Analysis························65
2.5 References···············································································································69
APPENDIX 1: NMR Spectra of Compounds Relevant to Chapter 2·······························75
CHAPTER 3. Chiral Biaryl N-Heterocyclic Carbene-Palladium Catalysts with Anagostic C-H∙∙∙Pd Interaction for Enantioselective Desymmetric C-N Cross-Coupling···············100
3.1 Introduction············································································································101
3.2 Results and Discussion····························································································105
3.2.1 Synthesis of Ligands and Catalysts···································································105
3.2.2 Catalytic Properties of ImPy-Pd Complexes······················································106
3.2.3 X-ray Crystallography····················································································110
3.2.4 Non-Covalent Interaction Plot Analysis····························································111
3.2.5 Electronic Properties for F-ImPy Ligands by IR and 77Se-NMR···························112
3.3 Conclusion·············································································································114
3.4 Experimental Section······························································································115
3.4.1 General Information·······················································································115
3.4.2 Synthesis of Ligands and Catalysts···································································116
3.4.2.1 Preparation of Chiral Amines··································································116
3.4.2.2 Preparation of Chiral Imidazo[1,5-a]pyridinium Salts································122
3.4.2.3 Preparation of ImPy-Metal Complexes and ImPy-Se Adducts·····················131
3.4.3 Substrate Synthesis and Compound Characterization·········································137
3.4.4 General Procedures for Desymmetric C-N Bond Formation································147
3.4.5 X-ray Crystallography Data for 3f and L8-Pd(allyl)Cl·······································163
3.4.6 Non-Covalent Interaction Plot Analysis····························································176
3.5 References·············································································································181
APPENDIX 2: NMR Spectra of Compounds Relevant to Chapter 3······························190
APPENDIX 3: The Cartesian Coordinates (Å) for the L8-Pd Complex as a Function of the Dihedral Angle Pd-C1-C7-H1················································································255
CURRICULUM VITAE·····························································································27