GIST Scholar
Not a member yet
30271 research outputs found
Sort by
Inhibition of serotonin-Htr2b signaling in skeletal muscle mitigates obesity-induced insulin resistance
Get PDFObesity-induced insulin resistance is a major cause of metabolic disorders, including type 2 diabetes mellitus. Although peripheral serotonin (5-hydroxytryptamine, 5-HT) has been implicated in energy balance and metabolism, its effect on skeletal muscle insulin sensitivity remains unclear. Here we identified the 5-HT receptor 2b (Htr2b) as a critical regulator of insulin sensitivity and energy metabolism in the skeletal muscle. Using genetic and pharmacological approaches, we showed that muscle-specific Tph1-knockout (Tph1 MKO) mice fed a high-fat diet exhibited reduced body weight, increased lean mass and improved glucose tolerance compared with wild-type mice. The pharmacological inhibition of Htr2b in myotubes reversed palmitate-induced insulin resistance and increased glycolytic activity. Moreover, muscle-specific HTR2b-knockout (HTR2b MKO) mice exhibited improved glucose uptake, insulin sensitivity and overall metabolic health under high-fat-diet-induced obesity. Mechanistically, both Tph1 MKO and Htr2b MKO mice showed increased phosphorylation of AKT and AMPK, indicating improved insulin sensitivity and energy metabolism in the skeletal muscle. These findings demonstrate that 5-HT-Htr2b signaling negatively regulates insulin sensitivity and energy metabolism in skeletal muscles, providing new insights into the role of peripheral serotonin in muscle metabolism and potential therapeutic targets for metabolic disorders.TRUEsciescopuskcikci_cand
The Role of Seed Endophytes in the Drought Tolerance of the Xerophytic Invasive Plant Lactuca serriola
No full textIn nature, plants form holobionts with a wide array of microbes, among which seed endophytes transmitted vertically to offspring emerge as crucial partners, particularly for invasive plants colonizing new habitats. These seed endophytes exhibit diverse plant growth-promoting traits that aid in plant establishment and growth across various environments, including drought conditions. Notably, drought-tolerant bacteria capable of producing specific material exopolysaccharides play a pivotal role in enhancing the soil water preservation and promoting plant growth under drought stress. However, the effectiveness of these plant growth-promoting bacteria often varies depending on the host species or genotype. Given its xerophytic nature and propensity for wind dispersal in arid seasons, the invasive alien plant Lactuca serriola would be thought to benefit from drought-tolerant seed endophytes, potentially bolstering its resilience to drought stress. Therefore, investigating the interaction between drought-tolerant seed endophytes and L. serriola promises to yield valuable insights into the role of seed endophytes in the invasiveness and adaptive mechanisms of invasive plants. This study aims to identify and characterize drought stress-alleviating seed endophytes from wild L. serriola and elucidate the underlying mechanisms. To achieve this goal, seed endophytes were isolated and identified, and drought-tolerant EPS-producing bacteria were selected using in vitro assays. Subsequently, the stress-alleviating effects of selected isolates were examined across various plant species, including Arabidopsis thaliana, Lactuca serriola, Lactuca sativa, Zea mays, and Camelina sativa. Furthermore, the study investigated genotype-specific responses by assessing five different L. serriola. The potential causes of host-specificity were explored, focusing on plant root exudate components. Finally, the stress-alleviating mechanisms of isolates was elucidated through transcriptome analysis. The results of this study revealed that out of 128 bacterial seed endophytes isolated from wild L. serriola, 42 representative strains exhibited at least one plant growth-promoting trait and drought tolerance. Among these, 12 isolates were selected for in planta assays based on their performance under drought conditions. Notably, Kosakonia cowanii GG1 demonstrated significant improvements in soil water contents and shoot growth under stress compared to uninfected control plants, suggesting the potential of seed endophytes as a novel strategy for invasive plants to thrive in challenging environments. Moreover, the study found that the infection of seed endophytes yielded different outcomes in various plant species. While several isolates positively impacted the plant and soil characteristics of L. serriola under drought, the effects varied depending on the L. serriola genotype. Pearson’s correlation coefficients indicated that rhiozosheath formation played the crucial role in the stress-alleviating effect of isolates. However, the isolates positively affected A. thaliana or L. serriola did not exhibit any stress-alleviating effect on three model plants. Analysis of organic acids within the root exudate revealed qualitative differences among plant species, suggesting their involvement in host-specific interactions. Additionally, K. cowanii GG1 and Erwinia tasmaniensis YJ6, which were effective in L. serriola P2 demonstrated differential responses to organic acid components. These results imply the importance of host species in plant-bacteria interaction and suggest that the root exudate would be a possible factor involved in host-specificity. Finally, the study unveiled that E. tasmaniensis YJ6 primed the L. serriola P2 for drought resistance under benign water conditions. The infected plants exhibited induced signal pathway of the stress-related hormone abscisic acid and accumulated osmoprotectants before the onset of stress, suggesting the potential of E. tasmaniensis YJ6 for biological priming against drought stress. In conclusion, this thesis underscores the pivotal role of seed endophytes in the adaptive mechanisms of L. serriola against drought stress. It highlights the significance of host-specific interactions and biological priming mechanisms in understanding plant-bacteria interactions in the context of environmental stress.DoctorABSTRACT i
CONTENTS iii
LIST OF TABLES vi
LIST OF FIGURES vii
Chapter 1. INTRODUCTION 1
1.1 Plant growth-promoting bacteria 2
1.2 Invasive plant and plant growth-promoting seed endophytic bacteria 6
1.3 Drought stress-ameliorating effect of plant-associated bacteria 8
1.4 Host specificity of plant growth-promoting bacteria 11
1.5 Study species: Lactuca serriola L 12
1.6 Objectives of the dissertation 14
References 17
Chapter 2. Isolation and identification of stress-ameliorating endophytic bacteria from the seeds of L. serriola 24
2.1 Introduction 25
2.2 Materials and Methods 28
2.2.1 Seed collection 28
2.2.2 Isolation and identification of L. serriola seed endophytes 28
2.2.3 Plant growth-promoting trait assay 30
2.2.4 Drought tolerance assay 30
2.2.5 Exopolysaccharide quantification 31
2.2.6 Effects of isolates on A. thaliana under drought 32
2.3 Results 34
2.3.1 Identification of L. serriola seed endophytes 34
2.3.2 Plant growth-promoting trait assay 36
2.3.3 Drought tolerance assay and exopolysaccharide quantification 36
2.3.4 Effects of isolates on A. thaliana under drought 41
2.4 Discussion 44
2.4.1 Seed endophytic bacteria from L. serriola wild population 44
2.4.2 Plant growth-promoting traits and drought-tolerance of isolated endophytes 46
2.4.3 Drought stress-ameliorating effects of endophytes on A. thaliana 46
2.4.4 Conclusions 47
References 48
Chapter 3. Confirmation of host specificity and preference of seed endophytic bacteria against various plant species 54
3.1 Introduction 55
3.2 Materials and Methods 58
3.2.1 GFP transformation of bacterial strain 58
3.2.2 Effects of bacterial infection to L. serriola under drought conditions 59
3.2.3 Visualization of GFP-transformed bacteria under a microscope 62
3.2.4 Effects of bacterial infection on model plants under drought conditions 63
3.2.5 Analysis of root exudate profile with HPLC-MS 64
3.2.6 Chemotaxis assay to root exudate components 66
3.3 Results 67
3.3.1 Effects of bacterial infection to L. serriola under drought condition 67
3.3.2 Visualization of GFP-transformed bacteria 75
3.3.3 Effects of bacterial infection on model plants under drought conditions 78
3.3.4 Analysis of root exudate profile 82
3.3.5 Chemotaxis assay to root exudate components 84
3.4 Discussion 87
3.4.1 Genotype-specific effects of drought-tolerant endophytes on L. serriola 87
3.4.2 Effects of drought-tolerant endophytes on model plant species 88
3.4.3 Chemotatic responses of drought stress-ameliorating endophytes against organic acids 89
3.4.4 Conclusions 90
References 91
Chapter 4. Investigation of drought stress-ameliorating mechanisms of E. tasmaniensis YJ6 through transcriptome analysis 95
4.1 Introduction 96
4.2 Materials and Methods 99
4.2.1 Sample collection and RNA extraction 99
4.2.2 Library construction and sequencing 100
4.2.3 Data-preprocessing, read mapping and assembly 101
4.2.4 Differential gene expression and pathway enrichment analysis 104
4.3 Results 106
4.3.1 Effects of E. tasmaniensis YJ6 treatments to L. serriola under drought condition 106
4.3.2 Gene expression profiles of L. serriola 108
4.3.3 Functional enrichment analysis based on the Kyoto Encyclopedia of Genes and Genomics pathway 118
4.3.4 Functional enrichment analysis based on the Gene Ontology term 133
4.3.5 Functional enrichment analysis based on the Gene Ontology term related to over-represented KEGG pathway 139
4.4 Discussion 144
4.4.1 Drought-induced gene expression changes in L. serriola 144
4.4.2 E. tasmaniensis YJ6-induced gene expression changes in L. serriola 145
4.4.3 Priming effect of E. tasmaniensis YJ6 on L. serriola through ABA signal transduction 149
4.4.4 Limitations of this study and conclusions 151
References 153
Chapter 5. CONCLUSIONS 159
5.1 Isolation and Identification of Stress-ameliorating Endophytic Bacteria from the Seeds of L. serriola 160
5.2 Confirmation of Host Specificity and Preference of Seed Endophytic Bacteria anainst Various Plant Species 160
5.3 Investigation of Drought Stress-ameliorating Mechanisms of E. tasmaniensis YJ6 through Transcriptome Analysis 161
5.4 Limitations 162
5.5 Conclusions 164
References 167
ACKNOWLEDGEMENTS 16
Functionalized thermal emitters for ferroelectric polymer-based solid-state cooling
No full text전 지구적인 이상기후와 생활 수준의 향상으로 냉방에 대한 수요가 급격히 증가하고 있다. 건물 내에서 냉방기기에 의한 소비 전력은 전체 소비 전력의 약 20%를 차지하며, 이 수요는 2050년까지 세 배 이상 증가할 것으로 예상한다. 전기열량 효과(electrocaloric effect)에 기반을 둔 고체 냉각 기술(solid-state refrigeration)은 기존의 증기 압축 냉방을 대체할 수 있는 유망한 대안으로, 높은 에너지 효율, 소형화 가능성, 냉매 가스를 사용하지 않는다는 장점이 있다. 그러나 열량 효과를 이용한 냉방 장치의 최대 성능은 열량 재료의 단열 온도 변화(adiabatic temperature change)로 제한되며, 특히 태양 복사 강도가 높은 야외 환경에서는 그 성능이 더욱 저하된다. 본 논문에서는 복사 방열판(radiative heat sink)을 통합하여 열역학 순환(thermodynamic cycle)이 확장된 고체 냉각 시스템을 제안한다. 먼저, 열 전도성이 향상된 복사 냉각기를 도입하여 고체 냉각 장치의 열 방출 효율을 높였으며, 광학적/열적 계산과 야외 실험을 통해 검증하였다. 이어서, 냉각과 난방이 모두 가능한 이중 동작을 구현하기 위해 각도 민감도가 낮은 유색 복사 냉각기(colored radiative cooler)를 개발하였다. 마지막으로, 가역적인 전기열량 순환과 복사열 관리의 결합을 통해 극한 환경에서도 안정적이고 효율적인 열 조절이 가능한 냉난방기에 대한 가능성을 제시하고, 앞으로 해결해야 할 기술적 과제에 대해 논의하였다.|The global demand for cooling is rapidly increasing due to intense heat waves and rising living standards. In buildings, cooling accounts for nearly 20% of electricity consumption, and this demand is projected to triple by 2050. Solid-state refrigeration based on the electrocaloric effect offers a promising alternative to conventional vapor-compression refrigeration, with advantages such as high energy efficiency, compactness, and elimination of gaseous refrigerants. However, its practical performance is limited by the adiabatic temperature change of caloric materials, particularly under outdoor conditions with high solar irradiation. This dissertation presents an enhanced solid-state cooling system incorporating a radiative heat sink. First, a thermally conductive radiative cooler is introduced to improve heat dissipation, validated through optical–thermal simulations and outdoor experiments. Next, a colored radiative cooler is developed to enable dual-mode operation for both cooling and heating with angular robustness. Finally, the integration of radiative thermal management with a reversible electrocaloric cycle is explored as a pathway toward robust and efficient thermoregulation.DoctorAbstract (English) i
Abstract (Korean) iii
List of Contents v
List of Tables vii
List of Figures viii
1 Introduction 1
1.1 Global demand for sustainable and high-efficiency cooling solutions 1
1.2 Solid-State Cooling Based on the Electrocaloric Effect 4
1.3 Radiative Cooling for Functionalized Applications 7
1.4 Ojective and Outline 10
2 Nonmetallic thermally conductive radiative coolers 13
2.1 Introduction 15
2.1.1 Thermal equilibrium equations 16
2.1.2 Thermal resistance modeling of radiative cooling 19
2.2 Nanofiber-based design for enhanced thermal conductivity 21
2.2.1 Exfoliation of hexagonal boron nitride (h-BN) into boron nitride
nanosheets (BNNSs) 21
2.2.2 Electrospun/spray-coated thermally conductive nanofibers 23
2.2.3 Thermal and optical characterization of ESRC 27
2.3 Polymer composite design with uniform filler dispersion 29
2.3.1 Ferroelectric polymer matrix embedded with BNNSs 29
2.3.2 Thermal and optical characterization of TCRC 38
2.4 Summary 45
3 Electrocaloric cooling with radiative thermal management 46
3.1 Introduction 48
3.1.1 Inherent limitations of caloric-effect-based solid-state cooling 50
3.1.2 Thermal management strategies for wearable devices 52
3.2 Electrocaloric cooling device with electrostatic actuation 54
– v –
3.2.1 Electrocaloric effect in relaxor ferroelectric polymers 54
3.2.2 Thermal analysis of electrocaloric cooling 62
3.3 Electrocaloric cooling system integrated with thermally conductive ra-
diative heat sink 66
3.3.1 Design and optimization of the radiative heat sink-integrated
electrocaloric (R-iEC) cooling system 66
3.3.2 Thermal and optical characterization of R-iEC 75
3.3.3 Thermal equilibrium equation for the R-iEC cooling system 84
3.4 Summary 87
4 Conclusion and future directions 89
4.1 Conclusion 89
4.2 Future directions 91
4.2.1 Dual mode of R-iEC system for the global energy saving 91
4.2.2 Spatially-segmented Colored radiative cooler for precise control
of solar absorption 100
4.2.3 Self-actuated electrocaloric cooling systems 101
References 102
Acknowledgements 118
– vi
A Canonical Domain Approach for Efficient Generalization of 3D Human Pose Estimation
No full text최근 딥러닝의 발전으로 3차원 인간 자세 추정(3D Human Pose Estimation, HPE) 의성능이크게향상되었다.그러나훈련도메인과목표도메인간의도메인차이로인한 성능저하는여전히일반화에있어주요한도전과제로남아있다.기존의일반화접근방 식인 도메인 일반화(Domain Generalization, DG)와 도메인 적응(Domain Adaptation, DA)은일반적으로광범위한데이터증강또는목표도메인에특화된적응을요구하므로, 추가적인 계산 비용과 시간 소모를 초래한다. 이러한 문제를 보다 효율적으로 해결하기 위해, 본 논문에서는 훈련와 목표 도메인 을 통합된 표준 도메인으로 변환하는 새로운 표준 도메인 접근 방식(canonical domain approach)을제안한다.이로써목표도메인에서의추가적인미세조정(fine-tuning)없이 도일반화가가능해진다.표준도메인구성을위해,본논문은 2D–3D자세간의일관성을 보장하고 자세 패턴을 단순화하여 리프팅 네트워크(lifting network)의 학습 효율을 향 상시키는 새로운 2D–3D 자세 매핑을 생성하는 자세 표준화(pose canonicalization)를 도입한다. 훈련 및 목표 도메인의 표준화는 다음과 같이 이루어진다: (1) 훈련 도메인 에서는 표준화된 2D–3D 자세 쌍을 사용하여 리프팅 네트워크를 학습시키고, (2) 목표 도메인에서는 추론 전 투영 기하(perspective projection)의 특성과 카메라 내부 파라미 터 정보를 활용하여 입력 2D 자세를 2D 표준화 과정을 통해 표준화한다. 이로써 학습된 네트워크를 목표 도메인에 대한 추가 학습 없이 바로 적용할 수 있다. 또한,본논문은 2D–3D자세패턴을복잡하게만드는요인중하나인 2D–3D크기모 호성(scale ambiguity)을해결하기위해표준도메인접근방식을확장한다.이를위해 3D 자세의크기를해당 2D자세의크기에정렬하는스케일표준화(scale canonicalization)를 도입하고,이를기존의자세표준화와결합하여스케일확장표준화과정(scale-extended canonicalization process)을 구성한다. 이러한 확장된 접근 방식은 표준 도메인 상에서 x 및 y 차원에서의 정확도를 크게 향상시키는 반면, 상대적인 깊이 모호성(relative depth ambiguity)이 해결되지 않아 z 차원에서는 성능이 저하되는 문제가 있다. 이를 보완하기 위해, 본 논문은 차원별 앙 상블(dimension-wise ensemble) 기법을 제안한다. 이 기법은 z 차원에서 우수한 성능을 보이는 자세 표준화 모델과 x, y 차원에서 효과적인 스케일 확장 표준화 모델의 예측을 결합하여 전반적인 성능을 향상시킨다. Human3.6M, Fit3D, MPI-INF-3DHP등다양한공개데이터셋과리프팅네트워크를 활용한 실험을 통해, 제안된 방법이 모델에 구애받지 않으며, 추가적인 데이터 증강이 나 도메인 적응 없이도 교차 도메인 일반화 성능을 효과적으로 향상시킴을 입증하였다. 특히, 3DHP 데이터셋에 대한 교차 도메인 평가에서 최고 성능을 달성하였다.|Recent advancements in deep learning have significantly improved the performance of 3D Human Pose Estimation (HPE). However, performance degradation caused by domain gaps between source and target domains remains a major challenge to gen- eralization. Conventional generalization approaches, such as Domain Generalization (DG) and Domain Adaptation (DA), often require extensive data augmentation or target domain-specific adaptation, resulting in additional computational cost and time consumption. To address this issue more efficiently, this dissertation proposes a novel canonical domain approach that transforms both the source and target domains into a unified canonical domain, thereby alleviating the need for additional fine-tuning in the tar- get domain. To construct the canonical domain, pose canonicalization is introduced to generate a novel 2D–3D pose mapping that ensures 2D–3D pose consistency and simplifies 2D–3D pose patterns, enabling more efficient training of lifting networks. Canonicalization of both domains is achieved as follows: (1) in the source domain, a lifting network is trained using canonical 2D–3D pose pairs; and (2) in the target do- main, input 2D poses are canonicalized prior to inference through a 2D canonicalization process which leverages the properties of perspective projection and known camera in- trinsics. As a result, the trained network can be directly applied to the target domain without additional fine-tuning. This dissertation further extends the canonical domain approach to address 2D–3D scale ambiguity, a factor that also complicates the 2D–3D pose patterns. Specifically, scale canonicalization is introduced, which aligns the scale of a 3D pose with that of its corresponding 2D pose. This is combined with the pose canonicalization to form a scale-extended canonicalization process. While this extended approach significantly improves accuracy in the x and y di- mensions in the canonical domain, it compromises performance in the z dimension due to unresolved relative depth ambiguity. To address this limitation, a dimension- wise ensemble method is proposed. This method combines predictions from the pose canonicalization model, which is effective in the z dimension, with those from the scale- extended canonicalization model, which performs well in the x and y dimensions. Experiments conducted using various lifting networks and publicly available datasets (e.g., Human3.6M, Fit3D, MPI-INF-3DHP) demonstrate that the proposed method is model-agnostic and effectively enhances cross-dataset generalization performance without requiring additional data augmentation or domain adaptation. In particu- lar, it achieves state-of-the-art performance in cross-dataset evaluations on the 3DHP dataset.DoctorAbstract (English) i
Abstract (Korean) iv
List of Contents vi
List of Tables ix
List of Figures xi
1 Introduction 1
1.1 Background and Problem Statement 1
1.1.1 Background 1
1.1.2 Problem Statement 3
1.2 Motivation and Overview of the Proposed Approach 4
1.2.1 Motivation 4
1.2.2 Overview of the Proposed Approach 5
1.3 Contributions 7
1.4 Dissertation Structure 8
2 Related Work 10
2.1 Deep Learning-based Human Pose Estimation 10
2.2 Generalization Capability in 3D Human Pose Estimation 11
2.2.1 Domain Generalization 11
2.2.2 Domain Adapatation 23
2.3 Canonicalization in 3D Human Pose Estimation 31
2.4 Scale Ambiguity in 3D HPE 48
3 Preliminary Background 49
3.1 Domain Gap in 3D Human Pose Estimation 49
3.2 Depth Ambiguity in 3D Human Pose Estimation 50
3.3 2D-to-3D Lifting Problem 51
3.4 Conventional 2D-3D Mapping and its Canonicalization Strategies 53
4 Canonical Domain Approach 56
4.1 Overview of Canonical Domain Approach 56
4.2 Pose Canonicalization 59
4.2.1 2D–3D Pose Inconsistency in Conventional 2D–3D Mapping 59
4.2.2 Pose Canonicalization Process 62
4.2.3 Effect of Canonical 2D–3D Pose Mapping Compared to Conventional Mapping 67
4.2.4 Target Domain Canonicalization 68
4.3 Scale-extended Canonicalization 73
4.3.1 2D–3D Scale Ambiguity 73
4.3.2 Scale Canonicalization Process 74
4.3.3 2D Input Residual Connection 76
4.4 Dimension-wise Model Ensemble 77
4.5 Training and Inference Process 78
4.5.1 Training Process 79
4.5.2 Inference Process 79
5 Experiments Settings 81
5.1 Dataset 81
5.2 Experiment Details 82
5.2.1 Experiment Type 82
5.2.2 Dataset Details 83
5.2.3 Methods 83
5.2.4 Metrics 84
5.2.5 Implementation Details 85
6 Result and Discussion 87
6.1 Cross-scenario Evaluation 88
6.2 Cross-dataset Evaluation 90
6.3 Qualitative Results 94
6.4 Evaluation on Additional Lifting Networks 95
6.5 Ablation Study 97
6.6 Comparison with Domain Generalization and Adaptation Methods 100
6.7 Computational Cost Analysis 101
6.8 Discussion on Additional Metrics (N-MPJPE, PA-MPJPE) 102
7 Limitation and Future Work 104
8 Conclusion 106
References 108
A Abbreviations 118
B More appendix 119
B.1 Derivation for Conventional and Canonical Pose Mapping 119
B.2 Computational Cost Analysis 121
B.2.1 2D Canonicalization Process 121
B.2.2 Inverse Canonicalization Process 122
B.2.3 Total FLOPs Summary 122
B.3 2.5D Pose Generation Process 123
B.4 Residual Errors between 2D-scaled Canonical 3D pose and Canonical 2D pose in X-Y Plane 124
Acknowledgements 12
Post-Processing Techniques for Forward and Inverse Monte Carlo Rendering
No full text물리 기반 렌더링은 사실적인 이미지를 생성하기 위해 널리 연구되어 왔으며, 최근에는 목표 이미지로부터 장면 매개변수를 기울기 기반 최적화 알고리즘을 기반으로 추정하는 역방향 렌더링 프레임워크에도 적용되고 있다. 이러한 프레임워크에서는 물리 기반 미분 가능한 렌더링을 통해 기울기를 계산한다. 이 방법들은 픽셀 색상과 그 도함수를 편향 없이 계산하기 위해 몬테카를로 기법을 사용하지만, 이로 인해 렌더링된 이미지와 기울기 모두에 노이즈(즉,몬테카를로분산)가 발생하게 된다.이러한 노이즈는 시각적으로 거슬리는 아티팩트를 유발할 수 있으며, 역방향 렌더링에서는 장면 매개변 수 최적화 알고리즘의 수렴을 방해할 수 있다. 정방향 렌더링의 노이즈를 줄이기 위해 다양한 이미지 디노이징 및 사후 보정 기술과 같은 후처리 기술들이 제안되어 왔다. 역방향 렌더링에서는 기존에 순방향 렌더링에서 제안된 이미지 디노이징 방식을적용하여 기울기의 노이즈를 줄이고, 이를 통해 최적화 성능을 개선하기 위한 연구가 수행되었다. 하지만, 이러한 접근법은 다양한 렌더링 상황에서 항상 이미지 품질을 개선하지 못하며, 역방향 렌더링에서는 디노이징 편향(bias)으로 인해 최적화가 정의되지 않은 국소 최소 값에 수렴하는 문제를 야기할 수 있다. 이러한 문제를 완화하기 위해 본 학위논문에서는 다양한 테스트 장면에서도 안정적으로 이미지 품질을 향상시키고, 디노이징 편향을 줄이며 역방향 렌더링의 최적화 수렴을 개선할 수 있는 후처리 기법들을 제안한다. 먼저, 제임스-스타인 추정기를 활용하여 편향이 없는 잡음(노이즈) 이미지와 편향이 있는 이미지(디노이징 결과)를 결합 함으로써 렌더링된 이미지의 품질을 향상시키는 사후 보정 기법을 제안한다. 이는 렌더링 분야에 제임스-스타인 추정기를 적용한 첫번째 연구이다. 역방향 몬테카를로 렌더링 에서는, 기존에 개발된 이미지 디노이징 기법의 직접적인 적용 대신 역방향 렌더링 특화 정보(즉, 목표 이미지)를 활용하여 더 강건하게 장면 매개변수 최적화를 개선할 수 있는 새로운 이미지 디노이징 방법을 제안한다. 마지막으로, 매개변수 공간에서 편향 없는 기울기와 편향된 기울기(예: 필터링된 기울기)를 제임스-스타인 추정기에 기반하여 국소적으로 결합하는 역방향 렌더링을 위한 사후 보정 기법을 제안 한다. 결합된 기울기는 최적화 알고리즘의 입력으로 사용되며 단일 유형의 기울기만을 사용하는 경우보다 더 빠른 수렴을 유도하는 것을 목표로 한다. 제안한 기법들의 결과로,정방향 렌더링을 위한 후처리 방법은 다양한 테스트 장면에서 일관되게 이미지 품질을 향상시키며, 최신 후처리 기법들보다 우수한 성능을 보였다. 또한, 역방향 렌더링을 위한 이미지 디노이징 및 사후 보정 기법은 최적화 알고리즘의 수렴을 개선함으로써 기존 기술들 보다 장면 매개변수를 더욱 정확하게 추정할 수 있음을 입증하였다.|Physically based rendering has been widely studied for generating photorealistic images and has recently been integrated into inverse rendering frameworks that estimate scene parameters from target images via gradient-based optimization. In these frameworks, gradients are computed through physically based differentiable rendering. These methods use Monte Carlo methods to compute pixel colors and their derivatives in an unbiased manner, but they introduce noise (i.e., Monte Carlo variance) in both rendered images and gradients. This noise can introduce visually distracting artifacts in rendered images and lead to slow convergence of scene parameter optimization in inverse rendering. In forward rendering, various image denoising and post-correction techniques have been proposed to reduce errors in rendered images. In inverse rendering, some of these denoisers have been directly adapted to generate less noisy gradients and improve the convergence of optimization. However, these approaches often struggle to consistently improve image quality across diverse rendering scenarios in forward rendering. In inverse rendering, these drive the optimization convergence to undesirable local minima due to denoising bias. To alleviate these issues, we propose post-processing techniques that robustly improve image quality across diverse test scenes and improve inverse rendering convergence with reduced denoising bias. First, we propose a post-correction technique that improves the quality of rendered images by combining unbiased and biased rendering estimates (e.g., a noisy image and its denoised output) using the James-Stein estimator. To the best of our knowledge, this is the first application of the James-Stein estimator in rendering. In inverse Monte Carlo rendering, we present a new image denoising method using inverse rendering-specific information (i.e., target images) to robustly improve the inverse rendering optimization compared to the direct adaptation of existing image denoising techniques. Finally, we propose a post-correction method that locally fuses unbiased and biased gradients (e.g., filtered gradients) in the parameter space, based on the James–Stein estimator. The combined gradients are then used as input to a gradient-based optimizer, leading to faster convergence compared to using either unbiased or biased gradients alone. Our results demonstrate that the proposed post-correction method for forward rendering consistently improves image quality across a wide range of test scenes, outperforming recent post-processing techniques. In addition, the proposed post-processing methods (image denoising and post-correction methods) enable more accurate inference of scene parameters through inverse rendering optimization than existing techniques.DoctorAbstract (English) i
Abstract (Korean) iii
List of Contents v
List of Tables vii
List of Figures viii
1 Introduction 1
1.1 (Forward) Monte Carlo Rendering 2
1.2 Inverse Monte Carlo Rendering 4
1.3 Proposed Methods for Forward and Inverse Rendering 5
1.4 Related Papers 6
2 Problem Definition and Related Work 7
2.1 Problem Definition 7
2.1.1 Monte Carlo Variance 7
2.1.2 Variance in Inverse Monte Carlo Rendering 9
2.2 Related Work 11
2.2.1 Monte Carlo Image Denoising 12
2.2.2 Post-Correction Techniques for MC Rendering 14
2.2.3 Differentiable Monte Carlo Rendering 15
3 Neural James-Stein Combiner for Unbiased and Biased Renderings 18
3.1 The James-Stein Estimator 19
3.2 Our Neural James-Stein Combiner 22
3.2.1 Localized James-Stein Combiner 22
3.2.2 Theoretical Discussion on the James-Stein Combiner 27
3.2.3 Optimization for the Localized James-Stein Combiner 29
3.2.4 Implementation of the Neural James-Stein Combiner 32
3.3 Results and Disccusions 33
3.4 The MSE of the Localized JS Combiner 48
– v –
4 Target-Aware Image Denoising for Inverse Monte Carlo Rendering 50
4.1 Image Denoising for Inverse Monte Carlo Rendering 51
4.2 Target-Aware Image Denoising 54
4.2.1 Bias Analysis of Our Denoised Image 57
4.2.2 Bias Analysis of Our Image-Space Gradient 60
4.3 Results and Discussion 61
4.3.1 Applications 63
4.3.2 Comparisons and Analysis 64
5 James-Stein Gradient Combiner for Inverse Monte Carlo Rendering 72
5.1 Unbiased and Biased Gradients for Inverse Monte Carlo Rendering 73
5.2 James-Stein Combiner with Unbiased and Biased Gradients 76
5.2.1 James-Stein Gradient Combiner 76
5.2.2 Adaptive Selection of Shrinkage Parameters 80
5.3 Results and Discussion 83
6 Conclusion 92
References 94
– vi
TeleHopper: Simulating a Jumping Sensation as Proprioceptive Feedback for Teleportation in Virtual Reality via Electrical Muscle Stimulation
No full textTeleportation, a method of instantly moving users to a target position, has become a widely adopted locomotion method in virtual reality. However, the lack of proprioceptive feedback for teleportation can diminish presence and increase workload, thereby limiting the overall user experience. In this study, we propose TeleHopper, a system that enhances the teleportation experience by simulating the sense of jumping during teleportation through Electrical Muscle Stimulation-based haptic feedback. TeleHopper induces leg movements resembling a jumping motion and adjusts stimulation intensity based on travel distance, creating a realistic proprioceptive perception of leaping through space during teleportation. Experimental results evaluating TeleHopper’s user experience showed a significant enhancement in sense of presence, as well as a significant reduction in mental workload. Through this study, we demonstrate TeleHopper’s ability to deliver compelling proprioceptive feedback in teleportation, with varying stimulation intensity enhancing realism and aiding travel distance estimation. © 2025 Copyright held by the owner/author(s)
Accelerator science in Korea: current challenges and future opportunities
No full textAccelerators that generate charged particle beams and boost them to the required energies have become an integral part of our daily life in various domains, ranging from materials processing and testing, medical diagnostics and therapy, and large-scale scientific discovery. This review traces the historical development of Korea's accelerator technology, summarizes the status of domestic large-scale accelerator facilities, and outlines strategic directions for next-generation infrastructures and applications. We highlight the current challenges and future opportunities across both discovery sciences and applied sectors, including global trends and Korea's position in the international landscape.FALSEsciescopuskc
Jung Ji-Yong’s Children’s Poetry: Perception of Reality and Strategies of Response
No full text이 논문은 정지용의 동시가 시적 실험과 자아 정립의 일환으로 출발하였으며 검열을 회피하기 위해 다양한 전략을 구사했다고 보고, 작품에 내재한 현실 인식을 분석하였다. ｢서쪽한울｣, ｢ᄯᅴ｣, ｢한울혼자보고｣는 폭력적인 세계를 비판적으로 형상화한 작품들이다. ｢서쪽한울｣은 석양에서 피, 화염, 전쟁을 감각하고, ｢ᄯᅴ｣는 띠가 상징하는 계층 질서의 허구성을 통찰하며, ｢한울혼자보고｣는 세계의 위협에 맞서 설화적 방어 기제를 구축한다. ｢감나무｣와 ｢넘어가는해｣는 수탈과 저항을 우화적으로 표현한 작품들이다. 홍시를 쪼아먹는 까마귀(｢감나무｣)와 들녘 지붕을 파먹는 불까마귀(｢넘어가는해｣)는 민족 자원을 수탈하는 외세를 상징한다. ｢감나무｣는 민요 ｢새 쫓는 소리｣의 전통에 기반하여 까마귀를 쫓아내는 적극적인 행동에 나서고, ｢넘어가는해｣는 사일(射日) 신화의 한문학 전고에 기대어 불까마귀가 쫓겨간 상황을 상상한다. ｢겨울ㅅ밤｣과 ｢해바락이씨｣는 사회적 실천의 가능성을 탐색하는 작품들이다. ｢겨울ㅅ밤｣은 미약한 빛과 소리를 통해 암흑기를 버티는 참여의 윤리를 보여주며, ｢해바락이씨｣는 동요운동에 참여하는 정지용의 문학적 실천 구상을 드러낸 작품이다. 이상의 분석을 통해 정지용의 동시가 아동문학인 동시에 식민지 현실에 대한 인식과 비판, 실천의 가능성을 내포한 시적 형식이었음을 살펴보았다.FALSEkc
Deep learning with guided attention for early diagnosis of Alzheimer's disease
No full textAlzheimer's Disease (AD) is one of the most common forms of neurodegenerative disease that involves the accumulation of amyloid beta plaques and tau tangles. The early diagnosis of AD is crucial as it helps patients to start preventive interventions to slow the disease's progression. We created a Guided-Attention Feature Extraction Deep Learning Network (GADL) for the early diagnosis of Alzheimer's disease (AD). We applied a GADL for the prediction of mild cognitive impairment (MCI) progression to AD and classification between MCI and cognitively normal (CN). We trained the model with magnetic resonance imaging images in the Alzheimer's Disease Neuroimaging Initiative (ADNI) database by subject-level data splitting and verified its generalizability in the Australian Imaging Biomarkers and Lifestyle Flagship Study of Aging (AIBL) database. Our method outperformed other subject-level studies with an accuracy of 80.29% for the prediction of MCI progression to AD and 83.70% for CN versus MCI classification in the ADNI dataset. The accuracies of our models when they were applied to the AIBL dataset are recorded as 79.38% and 79.83%, respectively. These results prove the high performance of our models in terms of its generalizability. The evaluation results showed that the proposed approach has competitive performance in comparison with recent studies in terms of its performance and generalizability. These results suggest that deep learning with guided attention can be an effective early diagnosis technique and a prognostic tool for Alzheimer's disease.TRUEsciescopu
