RISS 학술연구정보서비스

검색

인기 검색어

    다국어 입력

    http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.

    변환된 중국어를 복사하여 사용하시면 됩니다.

    예시)
    • 中文 을 입력하시려면 zhongwen을 입력하시고 space를누르시면됩니다.
    • 北京 을 입력하시려면 beijing을 입력하시고 space를 누르시면 됩니다.
    닫기

    Memory-Efficient Semantic Segmentation Networks for High-Resolution Remote Sensing Data = 고해상도 원격탐사 데이터를 위한 메모리 효율적인 의미 분할 네트워크

    한글로보기

    https://www.riss.kr/link?id=T17392798

    • 0

      상세조회
    • 0

      다운로드
    서지정보 열기
    • 내보내기
    • 내책장담기
    • 공유하기
      • URL 복사
    • 오류접수
    인용문이 복사되었습니다.

    부가정보

    목차 (Table of Contents)

    • I Introduction 1
    • 1.1 The Evolution of Semantic Segmentation and its Applications in Remote Sensing 1
    • 1.2 The Challenge of Very-High Resolution Remote Sensing Imagery 3
    • 1.3 The Imperative for Efficient Network Architectures 4
    • 1.4 Research Motivation and Problem Statement 6
    • I Introduction 1
    • 1.1 The Evolution of Semantic Segmentation and its Applications in Remote Sensing 1
    • 1.2 The Challenge of Very-High Resolution Remote Sensing Imagery 3
    • 1.3 The Imperative for Efficient Network Architectures 4
    • 1.4 Research Motivation and Problem Statement 6
    • 1.5 Thesis Contributions 8
    • 1.5.1 Mixer U-Net: Integrating ConvMixer and Inception Modules 8
    • 1.5.2 DASK-Net: Dual-Attention Selective Kernel Architecture 9
    • 1.5.3 LiDAF-Net: Lightweight Multimodal Fusion Architecture 11
    • 1.6 Thesis Structure and Organization 12
    • II Background and Related Work 15
    • 2.1 Introduction to Semantic Segmentation 15
    • 2.1.1 Classical Machine Learning Methods 15
    • 2.1.2 Deep Learning-Based Semantic Segmentation Methods 17
    • 2.1.3 Limitations and Contributions 21
    • 2.2 Lightweight Network Design 22
    • 2.2.1 Efficient Convolution Operations 22
    • 2.2.2 Efficient Architecture Design Strategies 23
    • 2.2.3 Lightweight Semantic Segmentation Networks 24
    • 2.2.4 Limitations and Contributions 27
    • 2.3 Attention Mechanisms for Semantic Segmentation 27
    • 2.3.1 Channel Attention 28
    • 2.3.2 Spatial Attention 28
    • 2.3.3 Dual-Attention Methods 29
    • 2.3.4 Selective Kernel Networks and Adaptive Receptive Fields 30
    • 2.3.5 Transformer-Based Self-Attention 31
    • 2.3.6 Limitations and Contributions 32
    • 2.4 Road Extraction from Remote Sensing Imagery 33
    • 2.4.1 Early Approaches 33
    • 2.4.2 Deep Learning-Based Road Extraction 34
    • 2.4.3 Limitations and Contributions 35
    • 2.5 Building Extraction from Remote Sensing Imagery 36
    • 2.5.1 Early Approaches 36
    • 2.5.2 Advanced Deep Learning Architectures 36
    • 2.5.3 Limitations and Contributions 37
    • 2.6 Multimodal Fusion for Semantic Segmentation 38
    • 2.6.1 Fusion Strategy Taxonomy 39
    • 2.6.2 Multimodal Fusion Architectures 40
    • 2.6.3 Limitations and Contributions 41
    • III Mixer U-Net: An Improved Automatic Road Extraction from UAV Imagery 43
    • 3.1 Introduction 43
    • 3.2 Preliminaries 48
    • 3.2.1 Building Blocks 48
    • 3.2.2 U-Net for Biomedical Image Segmentation 48
    • 3.2.3 ConvMixer Layer 53
    • 3.3 Proposed Model Architectures 55
    • 3.3.1 Mixer U-Net 55
    • 3.3.2 Mixer U-Net with Path Networks 57
    • 3.4 Experiments, Results and Discussions 58
    • 3.4.1 The Dataset and Preprocessing 58
    • 3.4.2 Implementation Details 59
    • 3.4.3 Baseline Models 60
    • 3.4.4 Evaluation Metrics 64
    • 3.4.5 Results and Analysis 65
    • 3.5 Conclusions 68
    • IV DASK-Net: A Lightweight Dual-Attention Selective Kernel Network for Efficient Dense Prediction in Remote Sensing Imagery 70
    • 4.1 Introduction 70
    • 4.2 Related works 75
    • 4.2.1 Road Network Extraction from Remote Sensing Imagery 76
    • 4.2.2 Building Extraction from Remote Sensing Imagery 77
    • 4.2.3 Attention Mechanisms and Multiscale Architectures 79
    • 4.2.4 Multimodal Semantic Segmentation methods 80
    • 4.3 Methodology 81
    • 4.3.1 Overview 82
    • 4.3.2 EI Unit 84
    • 4.3.3 Double Attention Selective Kernel module 86
    • 4.4 Experimental settings and results 93
    • 4.4.1 Experimental settings 93
    • 4.4.2 Results on Massachusetts roads dataset 98
    • 4.4.3 Results on the DeepGlobe dataset 101
    • 4.4.4 Results on WHU building dataset 102
    • 4.4.5 Ablation studies 105
    • 4.5 Conclusion and Future Works 111
    • V LiDAF-Net: A Lightweight Multi-Scale Depth-Aware Fusion Network for Multimodal Semantic Segmentation of High-Resolution Remote Sensing Imagery 113
    • 5.1 Introduction 113
    • 5.1.1 Background and Motivation 113
    • 5.1.2 Multimodal Fusion for Remote Sensing 114
    • 5.1.3 Contributions 116
    • 5.1.4 Chapter Organization 117
    • 5.2 Related Work 118
    • 5.2.1 Semantic Segmentation Networks 118
    • 5.2.2 Lightweight Network Design 119
    • 5.2.3 Multimodal Fusion for Semantic Segmentation 120
    • 5.2.4 Attention Mechanisms 121
    • 5.3 Methodology 122
    • 5.3.1 Overall Architecture 122
    • 5.3.2 Lightweight Fusion 124
    • 5.3.3 Depth-Aware Dual-Attention Selective Kernel Fusion 125
    • 5.3.4 Lightweight Attention Decoder 132
    • 5.4 Experimental Settings and Results 136
    • 5.4.1 Experimental Settings 136
    • 5.4.2 Results 139
    • 5.4.3 Ablation Studies 144
    • 5.5 Conclusion 148
    • 5.5.1 Broader Impact 149
    • VI Conclusion 150
    • 6.1 Summary of Contributions 150
    • 6.2 Key Insights and Lessons Learned 152
    • 6.3 Limitations and Open Challenges 155
    • 6.4 Future Research Directions 156
    • 6.5 Broader Impact and Concluding Remarks 158
    • References 161
    • Abstract (In English) 185
    • Abstract (In Korean) 187
    더보기

    분석정보

    View

    상세정보조회

    0

    Usage

    원문다운로드

    0

    대출신청

    0

    복사신청

    0

    EDDS신청

    0

    동일 주제 내 활용도 TOP

    더보기

    주제

    연도별 연구동향

    연도별 활용동향

    연관논문

    연구자 네트워크맵

    공동연구자 (7)

    유사연구자 (20) 활용도상위20명

    해외이동버튼