Spatial Temporal Graph Deconvolutional Network for Skeleton-Based Human Action Recognition

Benefited from the powerful ability of spatial temporal Graph Convolutional Networks (ST-GCNs), skeleton-based human action recognition has gained promising success. However, node interaction through message propagation does not always provide complementary information. Instead, it May even produce destructive noise and thus make learned representations indistinguishable. Inevitably, graph representation would also become over-smoothing especially when multiple GCN layers are stacked. This paper proposes spatial-temporal deconvolutional networks (ST-GDNs), a novel flexible deconvolution technique, to alleviate this issue. At its core, method provides better aggregation by removing embedding redundancy input graphs either node-wise, frame-wise or element-wise at different network layers. Extensive experiments on three current most challenging benchmarks verify that ST-GDN consistently improves performance largely reduce model size these datasets.