A Central Difference Graph Convolutional Operator for Skeleton-Based Action Recognition

This paper proposes a new graph convolutional operator called central difference convolution (CDGC) for skeleton based action recognition. It is not only able to aggregate node information like vanilla operation but also gradient information. Without introducing any additional parameters, CDGC can replace in existing Graph Convolutional Networks (GCNs). In addition, an accelerated version of the developed which greatly improves speed training. Experiments on two popular large-scale datasets NTU RGB+D 60 & 120 have demonstrated efficacy proposed CDGC. Code available at https://github.com/iesymiao/CD-GCN .