The key idea is to combine two powerful networks: a Spatial Graph Convolutional Network unit that understands intra-frame interactions to extract powerful features from different hand joints and a ...