簡易檢索 / 詳目顯示

回結果列表
研究生： 陳品翰
Chen, Pin-Han
論文名稱： 基於重要特徵之動畫生成
Keyframe Animation based on Important Features
指導教授： 李同益
Lee, Tong-Yee
學位類別： 碩士
Master
系所名稱： 電機資訊學院 - 資訊工程學系
Department of Computer Science and Information Engineering
論文出版年： 2020
畢業學年度： 108
語文別： 英文
論文頁數： 63
中文關鍵詞： 三維動畫 、隱含空間 、路徑探索 、形狀內插與重建 、As-Consistent-As-Possible(ACAP)
外文關鍵詞： 3D Animation, Latent Space, Path Exploration, Shape Interpolation and Reconstruction, As-Consistent-As-Possible(ACAP)
相關次數： 點閱：101  下載：0 
分享至:
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報

    • 本篇論文中，我們介紹一種流暢三維動畫的生成方法，使用者可以給定兩個或以上的三維模型，我們的系統將會自動生成流暢變化 的中間動作，這種方式不但能減少人工製作中間動作的過程，也能給予藝術家一些動作變化的參考。
    傳統的方法主要使用骨架來生成三維動畫，比起一張一張繪出動作，節省了很多時間，然而這種動畫的生成方式不一定保證產生合理的動作，其中一種可能的解決方式為：在骨骼附加一些特別的肌肉控制器。
    我們提出一種方法，透過含有大量合理動作的三維模型資料集，訓練一個卷積對抗自編碼器，來學習高維度資料空間中的流形，利用訓練好的編碼器將三維模型的資訊編碼成隱含空間中的低維度向量，給定這些低維度向量，我們利用貝茲曲線得出一條平滑的曲線，並在曲線上採樣節點，再利用解碼器將這些節點解碼回 ACAP特徵，再轉換回三維模型，希望能產生流暢的三維模型動畫。

    In this paper, we introduce a method for generating smooth 3D animations. The user could specify two or more 3D models, our system will automatically generate a continuously 3D deforming animation. This method not only reduces the time of manually making intermediate motions, but it can also give the artist inspiration for motion changes.
    The traditional method mainly uses the skeleton to generate three-dimensional animation, which saves a lot of time compared to creating 3D models one by one. However, the generation of this animation does not necessarily guarantee the generation of reasonable actions. One of the possible solutions is to attach some special muscle controllers to the bones.
    We propose a method to train a convolutional adversarial autoencoder using a 3D model dataset containing a large number of reasonable poses to learn the manifold of high dimensional data. We then use the trained encoder to encode the information of the three-dimensional model into low-dimensional vectors. Given these low-dimensional vectors, we adopt the concept of Bézier curve to get a smooth curve and sample nodes along the curve. Finally we use the trained decoder to decode these nodes back to ACAP (As-Consistent-As-Possible) features, and then convert them back to 3D models, hoping to produce a smooth 3D model animation .

    摘要 i Abstract ii 誌謝 iv Table of Contents v List of Tables vii List of Figures viii Chapter 1. Introduction 1 Chapter 2. Related Work 4 Shape Interpolation 4 Mesh Deformation 5 VAE 7 Manifold Learning 9 Chapter 3. System Overview 12 Chapter 4. Method 16 4.1 Datasets 16 4.2 Deformation Representation 18 4.2.1 Mesh to ACAP Feature 21 4.2.2 ACAP Feature to Mesh 23 4.3 Feature Preprocessing 25 4.4 Network Structure 27 4.4.1 Encoder 28 4.4.2 Decoder 29 4.4.3 Discriminator 29 4.4.4 Loss Function 30 4.5 Path Exploration 33 4.5.1 Encoding 33 4.5.2 KNN-MST Graph 34 4.5.3 Bézier Curve 35 4.5.4 Path Exploration without Extra keyframes 36 4.5.5 Path Exploration with Extra Keyframes 39 4.5.6 Get Mesh Sequence 40 Chapter 5. Experimental Results 41 5.1 Results of Path Exploration 41 5.1.1 Drawback of Linear Interpolation on Mesh Vertices 41 5.1.2 Results of Algorithm 2 42 5.1.3 Results of Algorithm 3 50 5.2 Training Details 51 5.3 Comparisons of System Architecture 52 5.3.1 Feature Representation 53 5.3.2 Network Architecture 54 5.4 Discussion 55 Chapter 6. Conclusion 57 6.1 Conclusions 57 6.2 Limitation and Future Works 57 References 58

    [1] O. K.-C. a. T. C.-L. a. C. H.-K. a. C.-O. D. a. L. T.-Y. Au, Skeleton extraction by mesh contraction, ACM transactions on graphics (TOG), 2008.
    [2] N. a. Z. R. a. P. L. a. C.-O. D. a. S. E. a. B. C. Fish, Image Morphing With Perceptual Constraints and STN Alignment, Wiley Online Library, 2020.
    [3] K. a. W. Y. a. L. D. a. C.-O. D. a. C. B. Aberman, Unpaired Motion Style Transfer from Video to Animation, arXiv preprint arXiv:2005.05751, 2020.
    [4] K. a. L. P. a. L. D. a. S.-H. O. a. C.-O. D. a. C. B. Aberman, Skeleton-Aware Networks for Deep Motion Retargeting, arXiv preprint arXiv:2005.05732, 2020.
    [5] D. a. S. J. a. K. T. a. J. T. Holden, Learning motion manifolds with convolutional autoencoders, SIGGRAPH Asia 2015 Technical Briefs, 2015.
    [6] L. a. L. Y.-K. a. Y. J. a. L.-X. Z. a. X. S. a. K. L. Gao, Sparse data driven mesh deformation, IEEE transactions on visualization and computer graphics, 2019.
    [7] A. a. S. J. a. J. N. a. G. I. a. F. B. Makhzani, Adversarial autoencoders, arXiv preprint arXiv:1511.05644, 2015.
    [8] H. a. K. A. a. T. F. P. Shao, The riemannian geometry of deep generative models, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops, 2018.
    [9] M. a. C.-O. D. a. L. D. Alexa, As-rigid-as-possible shape interpolation, Proceedings of the 27th annual conference on Computer graphics and interactive techniques, 2000.
    [10] H.-K. a. L. T.-Y. Chu, Multiresolution mean shift clustering algorithm for shape interpolation, IEEE transactions on visualization and computer graphics, 2009.
    [11] K. G. a. S. R. W. a. P. J. Der, Inverse kinematics for reduced deformable models, ACM Transactions on graphics (TOG), 2006.
    [12] R. W. a. Z. M. a. G. C. a. P. J. Sumner, Mesh-based inverse kinematics, ACM transactions on graphics (TOG), 2005.
    [13] R. W. a. P. J. Sumner, Deformation transfer for triangle meshes, ACM Transactions on graphics (TOG), 2004.
    [14] T.-Y. a. W. Y.-S. a. C. T.-G. Lee, Segmenting a deforming mesh into near-rigid components, Springer, 2006.
    [15] L. a. L. Y.-K. a. L. D. a. C. S.-Y. a. X. S. Gao, Efficient and flexible deformation representation for data-driven surface modeling, ACM Transactions on Graphics (TOG), 2016.
    [16] D. P. a. W. M. Kingma, Auto-encoding variational bayes, arXiv preprint arXiv:1312.6114, 2013.
    [17] Q. a. G. L. a. L. Y.-K. a. X. S. Tan, Variational autoencoders for deforming 3d mesh models, Proceedings of the IEEE conference on computer vision and pattern recognition, 2018.
    [18] S.-W. a. C. M. C. a. N. H. L. T. a. Y. C.-K. a. Y. S.-Y. a. L. S.-S. a. L. T.-Y. Zhang, Animation Video Resequencing with a Convolutional AutoEncoder, SIGGRAPH Asia 2019 Posters, 2019.
    [19] T. a. F. M. a. K. V. G. a. R. B. C. a. A. M. Groueix, 3d-coded: 3d correspondences by deep deformation, Proceedings of the European Conference on Computer Vision (ECCV), 2018.
    [20] M. a. M. N. a. R. J. a. P.-M. G. a. B. M. J. Loper, SMPL: A skinned multi-person linear model, ACM transactions on graphics (TOG), 2015.
    [21] G. a. R. J. a. M. X. a. M. N. a. B. M. J. a. L. I. a. S. C. Varol, Learning from synthetic humans, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2017.
    [22] S. a. K. A. a. J. D. W. a. B. M. J. Zuffi, 3D menagerie: Modeling the 3D shape and pose of animals, Proceedings of the IEEE conference on computer vision and pattern recognition, 2017.
    [23] Z.-H. a. W. Q. a. C. K. a. Z. J. Jiang, Disentangled representation learning for 3D face shape, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2019.
    [24] Q. a. Z. J. a. L. Y.-K. a. Z. J. a. C. J. Wu, Alive caricature from 2d to 3d, Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, 2018.
    [25] B. a. Z. J. a. C. J. a. Z. J. Jiang, Disentangled Human Body Embedding Based on Deep Hierarchical Neural Network, IEEE Transactions on Visualization and Computer Graphics, 2020.
    [26] C. R. a. S. H. a. M. K. a. G. L. J. Qi, Pointnet: Deep learning on point sets for 3d classification and segmentation, Proceedings of the IEEE conference on computer vision and pattern recognition, 2017.
    [27] C. R. a. Y. L. a. S. H. a. G. L. J. Qi, Pointnet++: Deep hierarchical feature learning on point sets in a metric space, Advances in neural information processing systems, 2017.
    [28] N. a. Z. Y. a. L. Z. a. F. Y. a. L. W. a. J. Y.-G. Wang, Pixel2mesh: Generating 3d mesh models from single rgb images, Proceedings of the European Conference on Computer Vision (ECCV), 2018.
    [29] M. Alexa, Merging polyhedral shapes with scattered features, Proceedings Shape Modeling International'99. International Conference on Shape Modeling and Applications, 1999.
    [30] M. Alexa, Linear combination of transformations, ACM Transactions on Graphics (TOG), 2002.
    [31] M. a. S. R. a. P. M. a. G. M. Botsch, Deformation transfer for detail-preserving surface editing, Vision, Modeling & Visualization, 2006.
    [32] D. a. S. A. a. L. D. Cohen-Or, Three-Dimensional Distance Field Metamorphosis, ACM transactions on graphics (TOG), 1998.
    [33] D. L. a. T. C. D. James, Skinning mesh animations, ACM Transactions on Graphics (TOG), 2005.
    [34] T. a. S. H. a. K. F. Kanai, Metamorphosis of arbitrary triangular meshes, IEEE Computer Graphics and Applications, 2000.
    [35] M. a. M. N. J. a. P. H. Kilian, Geometric modeling in shape space, ACM SIGGRAPH 2007 papers, 2007.
    [36] V. a. S. A. Kraevoy, Cross-parameterization and compatible remeshing of 3D models, ACM Transactions on Graphics (TOG), 2004.
    [37] Y. a. S. O. a. L. D. a. C.-O. D. Lipman, Linear rotation-invariant coordinates for meshes, ACM Transactions on Graphics (TOG), 2005.
    [38] N. J. a. G. L. J. a. P. M. Mitra, Partial and approximate symmetry detection for 3D geometry, ACM Transactions on Graphics (TOG), 2006.
    [39] S. I. a. H. J. K. Park, Capturing and animating skin deformation in human motion, ACM Transactions on Graphics (TOG), 2006.
    [40] S. a. Y. C. Schaefer, Example-based skeleton extraction, Symposium on Geometry Processing, 2007.
    [41] R. W. a. Z. M. a. G. C. a. P. J. Sumner, Mesh-based inverse kinematics, ACM transactions on graphics (TOG), 2005.
    [42] P. a. C. C. a. P. S. Veenstra, Spectral clustering using the kNN-MST similarity graph, 2016 8th Computer Science and Electronic Engineering (CEEC), 2016.
    [43] D. a. S. J. a. K. T. Holden, A deep learning framework for character motion synthesis and editing, ACM Transactions on Graphics (TOG), 2016.
    [44] A. a. P. D. Jacobson, libigl: prototyping geometry processing research in C++, SIGGRAPH Asia 2017 courses, 2017.

    下載圖示
    2025-09-01公開
    QR CODE