princeton-vl/RAFT

Remove Context

AMontiB opened this issue · 0 comments

Hi,

thank you for your incredible work. I want to ask what you mean when, in the paper, you say "train the network with the context network removed".

Specifically, what did you provide to the update module? Did you provide two zero tensors or two times the tensors of the feature encoder, or you modified the architecture of the update module?

I appreciate any help you can provide.

Best Regards,
Andrea