Paper Info
Title
3DViewGraph: Learning Global Features for 3D Shapes from A Graph of Unordered Views with Attention.
Abstract
Learning global features by aggregating information over multiple views has been shown to be effective for 3D shape analysis. For view aggregation in deep learning models, pooling has been applied extensively. However, pooling leads to a loss of the content within views, and the spatial relationship among views, which limits the discriminability of learned features. We propose 3DViewGraph to resolve this issue, which learns 3D global features by more effectively aggregating unordered views with attention. Specifically, unordered views taken around a shape are regarded as view nodes on a view graph. 3DViewGraph first learns a novel latent semantic mapping to project low-level view features into meaningful latent semantic embeddings in a lower dimensional space, which is spanned by latent semantic patterns. Then, the content and spatial information of each pair of view nodes are encoded by a novel spatial pattern correlation, where the correlation is computed among latent semantic patterns. Finally, all spatial pattern correlations are integrated with attention weights learned by a novel attention mechanism. This further increases the discriminability of learned features by highlighting the unordered view nodes with distinctive characteristics and depressing the ones with appearance ambiguity. We show that 3DViewGraph outperforms state-of-the-art methods under three large-scale benchmarks.
Year
DOI
Venue
2019
10.24963/ijcai.2019/107
IJCAI
Field
DocType
Volume
Common spatial pattern,Spatial analysis,Pattern recognition,Computer science,Pooling,Latent semantic mapping,Correlation,Artificial intelligence,Deep learning,Ambiguity,Shape analysis (digital geometry)
Journal
abs/1905.07503
Citations 
PageRank 
References 
7
0.45
0
Authors
6
Name
Order
Citations
PageRank
Han Zhizhong119818.28
Xiyang Wang2121.18
Chi-Man Vong355741.41
Yu-shen Liu431923.20
Zwicker Matthias52513129.25
C. L. Philip Chen64022244.76