HOW to use the graph similarity in traclets association 

I read your paper, and I think you input the GNN built trajectory features instead of the original trajectory average features into the clustering algorithm? Where is the relevant call code, and the code for the online