基于图像视频的三维建模是3D数字技术的核心内容，可以重建真实3D场景和人物，广泛应用于机器人及自动驾驶等领域，属于跨学科研究领域，具有很高的研究和应用价值。本书围绕图像视频三维建模的最新研究技术和方法展开，重点关注具有挑战性问题进行了系统研究和介绍，包括3D物体建模、3D人脸面部建模、3D人体姿态建模及通用建模的相关学习算法，是一本系统介绍三维建模先进方法的研究专著。本书中描述的所有算法都来自我们的研究成果，与最先进方法进行了比较，验证了有效性和先进性。本书将使人工智能及信息计算机等领域的研究人员、专业人士和研究生受益，对跨学科研究也非常有用。吴素萍，教授，一直从事计算机相关的教学科研工作，研究方向为机器视觉，人工智能，大数据和并行分布计算。担任中国计算机学会高性能计算专业委员。为国家人工智能，大数据重点研发项目评审专家。Chapter 1 ?Introduction11.1 ?3D Object Modeling21.1.1 ?Single_View 3D Reconstruction21.1.2 ?Multi_View 3D Reconstruction Method31.2 ?3D Face Modeling51.2.1 ?3D Face Keypoint Detection51.2.2 ?3D Face Reconstruction61.3 ?3D Human Body Modeling91.3.1 ?3D Human Pose Estimation91.3.2 ?3D human Body Reconstruction101.4 ?3D Reconstruction Modeling121.5 ?Outline of the Work13Bibliography14Chapter 2 ?3D Object Modeling172.1 ?Single_View 3D Object Modeling182.1.1 ?Multi_Scale Edge_Guided Learning for 3D Reconstruction182.1.2 ?Multi_Granularity Relationship Reasoning Network for High_Fidelity 3D Shape Reconstruction442.1.3 ?3D Shape Reconstruction Based on Dynamic Multi_Branch Information Fusion672.1.4 ?Hierarchical Feature Learning Network for 3D Object Reconstruction782.2 ?Multi_View 3D Object Modeling942.2.1 ?High_Resolution Multi_View Stereo with Dynamic Depth Edge Flow942.2.2 ?Global Contextual Complementary Network for Multi_View Stereo1032.2.3 ?Attention_Guided Multi_View Stereo Network for Depth Estimation1142.2.4 ?Self_Supervised Edge Structure Learning for Multi_View Stereo and Parallel Optimization1262.2.5 ?Layered Decoupled Complementary Networks for Multi_View Stereo1382.2.6 ?Global Balanced Networks for Multi_View Stereo149Bibliography157Chapter 3 ?3D Face Keypoint Detection1723.1 ?Learning Relation_Sensitive Structured Network for Robust Face Alignment1733.1.1 ?Introduction1733.1.2 ?Proposed Method1743.1.3 ?Experiments1783.1.4 ?Conclusion1813.2 ?Multi_Agent Deep Collaboration Learning for Face Alignment under Different Perspectives1823.2.1 ?Introduction1823.2.2 ?Proposed Method1843.2.3 ?Experiments1863.2.4 ?Conclusion1893.3 ?Towards Accurate 3D Face Alignment under Extreme Scenarios Via Multi_Granularity Perturbation Relearning1903.3.1 ?Introduction1903.3.2 ?Proposed Method1923.3.3 ?Loss Function1963.3.4 ?Experiments1973.3.5 ?Conclusion200Bibliography201Chapter 4 ?3D Face Reconstruction2054.1 ?Towards Rich_Detail 3D Face Reconstruction and Dense Alignment via Multi_Scale Detail Augmentation2064.1.1 ?Introduction2064.1.2 ?Proposed Method2074.1.3 ?Experiments2114.1.4 ?Conclusion2144.2 ?Multi_Attribute Regression Network for Face Reconstruction2154.2.1 ?Introduction2154.2.2 ?Proposed Method2174.2.3 ?Experiments2204.2.4 ?Conclusion2274.3 ?Geometry Normal Consistency Loss for 3D Face Reconstruction and Dense Alignment2284.3.1 ?Introduction2284.3.2 ?Proposed Method2304.3.3 ?Experiments2344.3.4 ?Conclusion2384.4 ?Complementary Learning Network for 3D Face Reconstruction and Alignment2384.4.1 ?Introduction2384.4.2 ?Proposed Method2404.4.3 ?Experiments2434.4.4 ?Conclusion2494.5 ?Graph Structure Reasoning Network for Face Alignment and Reconstruction2504.5.1 ?Introduction2504.5.2 ?Proposed Method2524.5.3 ?Experiments2564.5.4 ?Conclusion2604.6 ?Unsupervised Shape Enhancement and Factorization Machine Network for 3D Face Reconstruction2614.6.1 ?Introduction2614.6.2 ?Proposed Method2634.6.3 ?Experiments2664.6.4 ?Conclusion2704.7 ?A Detail Geometry Learning Network for High_Fidelity Face Reconstruction2714.7.1 ?Introduction2714.7.2 ?Proposed Method2734.7.3 ?Experiments2774.7.4 ?Conclusion2814.8 ?A Bi_Directional Optimization Network for De_Obscured 3D High_Fidelity Face Reconstruction2824.8.1 ?Introduction2824.8.2 ?Proposed Method2844.8.3 ?Experiments2894.8.4 ?Conclusion294Bibliography294Chapter 5 ?3D Human Pose Estimation3045.1 ?Multi_Hybrid Extractor Network for 3D Human Pose Estimation3055.1.1 ?Introduction3055.1.2 ?Proposed Method3065.1.3 ?Experiments3105.1.4 ?Conclusion3125.2 ?3D Human Pose Estimation Based on Center of Gravity3125.2.1 ?Introduction3125.2.2 ?Proposed Method3155.2.3 ?Experiments3195.2.4 ?Conclusion3245.3 ?Edge_Angle Structure Constraint Loss for 3D Human Pose Estimation3245.3.1 ?Introduction3245.3.2 ?Related Works3255.3.3 ?Proposed Method3265.3.4 ?Experiments3295.3.5 ?Conclusion333Bibliography333Chapter 6 ?3D Human Body Reconstruction3396.1 ?Two_Stage Co_Segmentation Network Based on Discriminative Representation for Recovering Human Mesh from Videos3406.1.1 ?Introduction3406.1.2 ?Related Works3426.1.3 ?Proposed Method3436.1.4 ?Experiments3506.1.5 ?Conclusion3566.2 ?Frame_Level Feature Tokenization Learning for Human Body Pose and Shape Estimation3566.2.1 ?Introduction3566.2.2 ?Related Works3586.2.3 ?Proposed Method3596.2.4 ?Experiments3636.2.5 ?Conclusion3686.3 ?Time_Frequency Awareness Network for Human Mesh Recovery from Videos3696.3.1 ?Introduction and Related Works3696.3.2 ?Proposed Method3716.3.3 ?Experiments3756.3.4 ?Conclusion3786.4 ?Spatio_Temporal Tendency Reasoning for Human Body Pose and Shape Estimation from Videos3786.4.1 ?Introduction and Related Works3786.4.2 ?Proposed Method3806.4.3 ?Experiments3856.4.4 ?Conclusion389Bibliography389Chapter 7 ?3D Reconstruction Modeling3947.1 ?Replay Attention and Data Augmentation Network for 3D Face and Object Reconstruction3957.1.1 ?Introduction3957.1.2 ?Related Works3987.1.3 ?Proposed Method4027.1.4 ?Experiments4087.1.5 ?Conclusion4177.2 ?A Lightweight Grouped Low_Rank Tensor Approximation Network for 3D Mesh Reconstruction from Videos4177.2.1 ?Introduction4177.2.2 ?Proposed Method4207.2.3 ?Experiments4237.2.4 ?Conclusion429Bibliography429