This repository provides training and evaluation code for paper titled "Graph Convolutional Networks for Assessment of Physical Rehabilitation Exercises" (accepted in IEEE TNSRE)
Self attention was calculated using convlstm network which calculate spatial and temopral attention of each joint. You can found the code description in sgcn.py file.
I want to know where the self-attention was added, and the self-attention matrix graph, is it calculated using the bias matrix