Application question

[zhang-qiang-github]

Dear tkipf, I have a question in our current project, and I need your help.

For example, there is a human being consisted of six lines (a matchstick men), and I need to detect the head, hand and the joint. This matchstick men may have different pose, the length of arm and leg may be varied, or the arm/leg may be missed. I want to use the graph neural network to solve this problem. However, in my application, one matchstick men is a sample which is different from that in graph neural network. Is my understanding correct?

For this application, do you have any suggestion? Thank you for your time.