Progressive Spatio-Temporal Graph Convolutional Network for Skeleton-Based Human Action Recognition

Graph convolutional networks (GCNs) have been very successful in\nskeleton-based human action recognition where the sequence of skeletons is\nmodeled as a graph. However, most of the GCN-based methods in this area train a\ndeep feed-forward network with a fixed topology that leads to high\ncomputational complexity and restricts their application in low computation\nscenarios. In this paper, we propose a method to automatically find a compact\nand problem-specific topology for spatio-temporal graph convolutional networks\nin a progressive manner. Experimental results on two widely used datasets for\nskeleton-based human action recognition indicate that the proposed method has\ncompetitive or even better classification performance compared to the\nstate-of-the-art methods with much lower computational complexity.\n