search

kenziyuliu / DGNN-PyTorch

Unofficial PyTorch implementation of the CVPR'19 paper "Skeleton-Based Action Recognition with Directed Graph Neural Networks".
Other
235 stars 59 forks source link

Not really adaptive graph #19

Closed Scaramouch33 closed 4 years ago

Scaramouch33 commented 4 years ago

Recently, I have go through the code and find that in this implementation, the number of the edges is still fixed. But if we input all the possible edges into the network of 9 layers, the memory will boom. I wonder whether there are some good ideas to use all possible edges and can stack many layers too.

kenziyuliu commented 4 years ago

Hi,

I think by "adaptive" the paper means that the edge weights in the graph are learnable through backprop. This line allows gradients for the graph: https://github.com/kenziyuliu/DGNN-PyTorch/blob/master/model/dgnn.py#L64

15762260991 commented 1 year ago

Hi,

I think by "adaptive" the paper means that the edge weights in the graph are learnable through backprop. This line allows gradients for the graph: https://github.com/kenziyuliu/DGNN-PyTorch/blob/master/model/dgnn.py#L64

@kenziyuliu Is the same principle of "adaptive" in MS-G3D? Looking forward to your reply！

kenziyuliu commented 1 year ago

Hi @15762260991 this line in the MS-G3D repo might be relevant: https://github.com/kenziyuliu/MS-G3D/blob/master/model/ms_gcn.py#L40

15762260991 commented 1 year ago

Hi @15762260991 this line in the MS-G3D repo might be relevant: https://github.com/kenziyuliu/MS-G3D/blob/master/model/ms_gcn.py#L40

Thank you for your patience and guidance!