Point2Skeleton: Learning Skeletal Representations from Point Clouds 
Cheng Lin^{1}, Changjian Li^{2}, Yuan Liu^{1}, Nenglun Chen^{1}, YiKing Choi^{1}, Wenping Wang^{1} 
^{1}The University of Hong Kong, ^{2}University College London 
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2021. 
Oral Presentation 


Fig. 1. We introduce an unsupervised method to learn skeletal meshes from point clouds. The skeletal meshes contain both 1D curve segments and 2D surface sheets which can represent underlying structures of various shapes.

Abstract 
We introduce Point2Skeleton, an unsupervised method to learn skeletal representations from point clouds. Existing skeletonization methods are limited to tubular shapes and the stringent requirement of watertight input, while our method aims to produce more generalized skeletal representations for complex structures and handle point clouds. Our key idea is to use the insights of the medial axis transform (MAT) to capture the intrinsic geometric and topological natures of the original input points. We first predict a set of skeletal points by learning a geometric transformation, and then analyze the connectivity of the skeletal points to form skeletal mesh structures. Extensive evaluations and comparisons show our method has superior performance and robustness. The learned skeletal representation will benefit several unsupervised tasks for point clouds, such as surface reconstruction and segmentation. 


Paper [PDF]
Code [Github]
Citation: Lin C, Li C, Liu Y, et al. Point2Skeleton: Learning Skeletal Representations from Point Clouds[J]. arXiv preprint arXiv:2012.00230, 2020. (bibtex)



Network Overview 

Fig. 2. An overview of our Point2Skeleton pipeline. Given a point cloud as input, we first learn a geometric transformation via convex combinations to predict the skeletal points together with their radii. Second, we connect the skeletal points to form mesh structures. We initialize a graph structure using two simple priors derived from the properties of skeletal mesh; and then we formulate a link prediction problem using a graph autoencoder to obtain a complete skeletal mesh.


Representative Results 
Comparison 

Fig. 3. Qualitative comparison with the competitive point cloud skeletonization methods, i.e., L1medial skeleton and deep point consolidation (DPC) 
Applications  Reconstruction 

Fig. 4. Unsupervised surface reconstruction from point clouds. Our method can produce watertight surfaces that capture more detail of thin structures without the need of normal vectors. 
Applications  Segmentation 

Fig. 5. Unsupervised structural decomposition for point clouds using the skeletal meshes generated by our method. 
Applications  Inpaiting 

Fig. 6. Skeletal mesh prediction for reconstructing complete geometries from point clouds with missing regions. 


©Changjian Li. Last update: March 7, 2021. 