TY - GEN
T1 - ROTATION-EQUIVARIANT GRAPH CONVOLUTIONAL NETWORKS FOR SPHERICAL DATA VIA GLOBAL-LOCAL ATTENTION
AU - Xu, Jiayi
AU - Yang, Qin
AU - Li, Chenglin
AU - Zou, Junni
AU - Xiong, Hongkai
AU - Pan, Xinlong
AU - Wang, Haipeng
N1 - Publisher Copyright:
© 2022 IEEE.
PY - 2022
Y1 - 2022
N2 - Graph convolutional networks (GCNs) are widely adopted for spherical data processing, striking a balance between rotation equivariance and computation efficiency. However, current GCN-based approaches either apply radial filters with a fixed convolution kernel, or adopt attention schemes to learn a variable kernel based only on node features, which lack expressiveness and may suffer over-smoothing problem. To address this, in this paper, we propose a global-local attention-based spherical graph convolution (GlasGC) for spherical data representation. Specifically, the designed graph convolution includes a structure-enhanced local attention module to estimate the structure correlation efficiently based on topological characteristics of the given spherical graph. The structure correlation is then incorporated with the feature correlation to form the local attention that determines importance of a neighbor node to the center node. We also design a global attention module to obtain the importance scores of different nodes based on both node features and graph topology, which acts as a message-passing enhancer for informative nodes and can relieve the over-smoothing problem. Based on our proposed GlasGC, we design specific GlasGCNs for classification and semantic segmentation of spherical data. Empirical evaluations on these tasks demonstrate that our proposed GlasGCNs can generally achieve better classification and segmentation performance.
AB - Graph convolutional networks (GCNs) are widely adopted for spherical data processing, striking a balance between rotation equivariance and computation efficiency. However, current GCN-based approaches either apply radial filters with a fixed convolution kernel, or adopt attention schemes to learn a variable kernel based only on node features, which lack expressiveness and may suffer over-smoothing problem. To address this, in this paper, we propose a global-local attention-based spherical graph convolution (GlasGC) for spherical data representation. Specifically, the designed graph convolution includes a structure-enhanced local attention module to estimate the structure correlation efficiently based on topological characteristics of the given spherical graph. The structure correlation is then incorporated with the feature correlation to form the local attention that determines importance of a neighbor node to the center node. We also design a global attention module to obtain the importance scores of different nodes based on both node features and graph topology, which acts as a message-passing enhancer for informative nodes and can relieve the over-smoothing problem. Based on our proposed GlasGC, we design specific GlasGCNs for classification and semantic segmentation of spherical data. Empirical evaluations on these tasks demonstrate that our proposed GlasGCNs can generally achieve better classification and segmentation performance.
KW - attention mechanism
KW - classification
KW - graph convolutional networks
KW - semantic segmentation
KW - Spherical images
UR - https://www.scopus.com/pages/publications/85146735825
U2 - 10.1109/ICIP46576.2022.9897510
DO - 10.1109/ICIP46576.2022.9897510
M3 - 会议稿件
AN - SCOPUS:85146735825
T3 - Proceedings - International Conference on Image Processing, ICIP
SP - 2501
EP - 2505
BT - 2022 IEEE International Conference on Image Processing, ICIP 2022 - Proceedings
PB - IEEE Computer Society
T2 - 29th IEEE International Conference on Image Processing, ICIP 2022
Y2 - 16 October 2022 through 19 October 2022
ER -