文字雲的出現，不論在學界或是業界都帶來新穎與實用的成效；相對文字，影像可表達更豐富的資訊，興起諸多以影像取代文字來創作更具特色的影像雲資訊視覺化研究。傳統上影像雲中的物件多以幾何形狀如矩形或菱形，以併排或相互重疊的方式拼貼於一簡單外型內，所能表達的層次較簡單。本研究的影像形狀雲拼貼則挑戰使用不規則的影像雲物件，拼貼於一個不規則且有關聯意涵的外型中，達到裡外呼應的視覺化效果；本研究更進一步整合時間軸的演進模擬，藉由生動的影像時變動態行為演示，將以更具親和力的方式呈現影像雲主題內容的視覺傳達(Visual Communication)。
在影像形狀雲拼貼上，本研究提出新的深度學習網路StuffNet進行shape matching。利用影像物件的輪廓搜尋可放置空間後，經由StuffNet進行shape matching，找出關聯矩陣對影像物件進行影像形狀仿射轉換(Affine Transformation)，使影像物件更貼合於可放置空間內，優化空間的使用率，減少空隙空間(empty space)。此外，StuffNet輸入的是原影像物件輪廓圖片，有別於傳統shape matching方法需要事先對影像物件輪廓曲線參數化，建立描述器(Descriptor)等冗長數值運算，相對此利用描述器之研究，本研究大大減少計算時間。在時變形狀雲模擬上，本研究提出一個完整的動態行為模擬系統。隨著時間的演進，它會建立動態互動圖(Dynamic Interaction Graph) 來應對每個時間步驟下可變的影像物件數、位置、相互關係，再藉由本研究以深度學習網路搭建的神經動態模擬器(Neural Dynamic Simulator, NDS)來執行物理運動模擬，在保持外型的同時連續平順地完成影像形狀雲之間的轉變，進而平順地完成整個時變模擬動態形狀雲轉變。
總而言之，本研究不論在影像形狀拼貼，或是影像時變模擬均擴展先前方法對不規則形狀影像物件應用的限制，同時融入了深度學習技術，而不再沿襲傳統數值分析與推論做法，使得在影像形狀雲時變的研究結果有所突破，於此應用的研究課題上能夠另闢新徑。

The emergence of word cloud has brought novelty and practicality to both academia and the industry. Compared to text, images can express richer information, which has led to the creation of more distinctive visualization studies of image clouds by replacing texts with images. Traditionally, the objects in image clouds are mostly geometric shapes such as rectangles or rhombuses, and they are collaged or overlapped to form a simple shape. This study challenges the use of irregular image shape objects in an arbitrary and associative shape to achieve the visualization effect of echoing the inside and outside. Our research further integrates the simulation of the timeline to present the visual communication of the time varying shape cloud in a more approachable way through the temporal changes of the image.
In this study, we propose a new deep learning network, StuffNet, for shape matching. To achieve the approach of image shape cloud, we first search all the placeable space by the contour of the shape. Second, performs shape matching by StuffNet, to find the correlation matrix for affine transformation, so that the transformed image shape can fit better in the space also minimize the gap between objects while not overlapping. In addition, the input of StuffNet is the original shape contour, which is different from the traditional shape matching method that needs to parameterize the shape contour or curves and build a shape descriptor beforehand.
For the time varying shape cloud, we proposes a dynamic behavior simulation system. It builds a dynamic interaction map to record object numbers, object position, etc. every time step. Using the Neural Dynamic Simulator (NDS) built by the deep learning network to perform physical simulation. Smoothly complete the transition between image shapes and clouds while maintaining the appearance.
In summary, this study break the limits of previous methods for the application of irregular image objects in both image shape collage and image time-varying simulation. Incorporating Deep learning techniques instead of following the traditional numerical analysis and inference approach. It has made breakthrough in the research of both shape cloud and time-varying image shape clouds and has opened up new avenues in the research of this application.

M. Jaderberg, K. Simonyan, A. Zisserman and K. Kavukcuoglu, "Spatial Transformer Networks," Computer Vision and Pattern Recognition, pp. 2017-2025, 2015.
K.-C. Kwan, J. Lok-Tsun Sinn, C. Han, T.-T. Wong and C.-W. Fu, "Pyramid of Arclength Descriptor for Generating Collage of Shapes," ACM Transactions on Graphics, pp. 229:1--229:12, 11 2016.
M.-T. Chi, S.-S. Lin, S.-Y. Chen, C.-H. Lin and T.-Y. Lee, "Morphable Word Clouds for Time-Varying Text Data Visualization," IEEE Transactions on Visualization and Computer Graphics, pp. 1415-1426, 12 2015.
F. B. Viegas, M. Watternberg and J. Feinberg, "Participatory Visualization with Wordle," IEEE Transactions on Visualization and Computer Graphics, pp. 1137-1144, 11 2009.
H. Strobelt, M. Spicker, A. Stoffel, D. Keim and O. Deussen, "Rolled-outWordles: A Heuristic Method for Overlap Removal of 2D Data Representatives," Comput. Graph. Forum, pp. 1135-1144, 2012.
K. Koh, B. Lee, B. Kim and J. Seo, "ManiWordle: Providing Flexible Control over Wordle," IEEE Transactions on Visualization and Computer Graphics, pp. 1190-1197, 11-12 2010.
R. Maharik, M. Bessmeltsev, A. Sheffer, A. Shamir and N. Carr, "Digital Micrography," ACM Trans. Graph, pp. 100:1-100:12, 12 2011.
Y. Wang, X. Chu, K. Zhang, C. Bao, X. Li, J. Zhang, C.-W. Fu, C. Hurter, O. Deussen and B. Lee, "ShapeWordle: Tailoring Wordles using Shape-aware Archimedean Spirals," IEEE Transactions on Visualization and Computer Graphics, pp. 991-1000, 1 2020.
W. Cui, Y. Wu, S. Liu, F. Wei, M. X. Zhou and H. Qu, "Context Preserving Dynamic Word Cloud Visualization," IEEE Comput. Graph. Appl., pp. 42-53, 11-12 2010.
T. Liu, J. Wang, J. Sun, N. Zheng, X. Tang and H. Shum, "Picture Collage," IEEE Transactions on Multimedia, pp. 1225-1239, 11 2009.
M. H. Lee, N. Singhal, S. Cho and I. K. Park, "Mobile photo collage," IEEE Computer Society Conference on Computer Vision and Pattern Recognition, pp. 24-30, 2010.
X. Han, Z. Chongyang, W. Lin, M. Xu, B. Sheng and T. Mei, "Tree-Based Visualization and Optimization for Image Collection," IEEE Transactions on Cybernetics, 7 2015.
Y.-M. Kuo, H.-K. Chu, M.-T. Chi, R.-R. Lee 且 T.-Y. Lee, “Generating Ambiguous Figure-Ground Images,” IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, pp. 1534-1545, 5 2017.
R. A. Saputra, C. S. Kaplan and P. Asente, "Improved Deformation-Driven Element Packing with RepulsionPak," IEEE Transactions on Visualization and Computer Graphics, pp. 2396-2408, 4 2021.
G. Pai, A. Wetzler and R. Kimmel, "Learning Invariant Representations Of Planar Curves," International Conference on Learning Representations, 2017.
A. Kanazawa, D. Jacobs and M. Chandraker, "WarpNet: Weakly Supervised Matching for Single-View Reconstruction," IEEE Conference on Computer Vision and Pattern Recognition, pp. 3253-3261, 2016.
R. Hanocka, N. Fish, Z. Wang, R. Giryes, S. Fleishman and D. Cohen-Or, "ALIGNet: Partial-Shape Agnostic Alignment via Unsupervised Learning," ACM Transactions on Graphics, 4 2018.
M. B. Chang, T. Ullman, A. Torralba and J. B. Tenenbaum, "A Compositional Object-Based Approach to Learning Physical Dynamics," International Conference on Learning Representations, 2017.
P. W. Battaglia, R. Pascanu, M. Lai, D. Rezende and K. Kavukcuoglu, "Interaction Networks for Learning about Objects, Relations and Physics," NIPS, 2016.
Y. Li, J. Wu, J.-Y. Zhu, J. B. Tenenbaum, A. Torralba and R. Tedrake, "Propagation Networks for Model-Based Control Under Partial Observation," International Conference on Robotics and Automation, 2019.
O. Ronneberger, P. Fischer and T. Brox, "U-Net: Convolutional Networks for Biomedical Image Segmentation," "Medical Image Computing and Computer-Assisted Intervention -- MICCAI 2015, pp. 234--241, 2015.
Y.-H. Tsai and T.-Y. Lee, "Image Shape Collage and Time-varying Image Shape Cloud Visualization," Unpublished doctoral dissertation, National Cheng Kung University, College of Electrical Engineering & Computer Science / Institute of Computer Science and Information Engineering, 2018.
L. J. Latecki, R. Lakamper and a. T. Eckhardt, "Shape descriptors for non-rigid shapes with a single," CVPR, p. 424–429, 2000.

http://mingtc.com/2017/06/20/%E8%A3%BD%E4%BD%9C%E6%96%87%
E5%AD%97%E9%9B%B2%E8%88%87%E4%B8%AD%E6%96%87%E7
%B7%9A%E4%B8%8A%E6%96%B7%E8%A9%9E/
https://www.google.com/url?sa=i&url=https%3A%2F%2Fkknews.cc%2Fne
ws%2Flxl6alz.html&psig=AOvVaw2-
BqbC2l9zycTI272H7E23&ust=1605519146612000&source=images&cd=vf
e&ved=0CAIQjRxqFwoTCPjgppCfhO0CFQAAAAAdAAAAABAJ
https://www.google.com/search?q=image+collage&sxsrf=ALeKk00d1WfLg
e34_i497CTyWlIS7dPwoQ:1620550991787&source=lnms&tbm=isch&sa=
X&ved=2ahUKEwi4hdLxnrzwAhVR62EKHfekDl4Q_AUoAXoECAEQAw
&biw=1920&bih=937#imgrc=kn4Jo-WUEppt0M
https://www.google.com/search?q=image+collage&sxsrf=ALeKk00d1WfLg
e34_i497CTyWlIS7dPwoQ:1620550991787&source=lnms&tbm=isch&sa=
X&ved=2ahUKEwi4hdLxnrzwAhVR62EKHfekDl4Q_AUoAXoECAEQAw
&biw=1920&bih=937#imgrc=gN7hSu51xHpKmM
https://box2d.org/
https://chipmunk-physics.net/
https://pybullet.org/wordpress/
http://www.pymunk.org/en/latest/