princeton-vl / RAFT

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

How about use deconvolution instead of convex upsampling in paper?

ducheng678 opened this issue · comments

I am curious about this as well, did you get an answer for this?