Fig. 4From: STTG-net: a Spatio-temporal network for human motion prediction based on transformer and graph convolution networkThe prediction revision module. The first line of ‘prediction’ is the result directly predicted by the network, while the second line is the “final prediction.” From the second frame, the current ‘prediction’ result is fused with the “final prediction” result of the previous frames and then calculated as the final prediction for the current frameBack to article page