search

wei-mao-2019 / LearnTrajDep

code for learning trajectory dependencies for human motion prediction
MIT License
244 stars 45 forks source link

Ablation Study #10

Closed catherinezll95 closed 3 years ago

catherinezll95 commented 4 years ago

Thanks for your code sharing. In Ablation Study, the paper mentioned the experiment of using a pre-defined adjacency matrix based on the kinematic tree. In the code, I found that you regard the position(x, y, z) of one node as three nodes in your graph. So I am curious about when you use a pre-defined adjacency matrix, you regard (x, y, z) as the attribute rather than three nodes? It means that the input nodes are 22 rather than 66 in your 3D training.

wei-mao-2019 commented 4 years ago

Hi @catherinezll95,

I use 66 nodes thus (x, y, z) as 3 nodes.

Thank you