search

vt-vl-lab / FGVC

[ECCV 2020] Flow-edge Guided Video Completion
Other
1.55k stars 263 forks source link

can you elaborate what is RAFT ? what does it stand for ? I did not see those initials in the paper #28

Closed dagnichz closed 3 years ago

gaochen315 commented 3 years ago

RAFT is a recent method (ECCV2020) for optical flow prediction. When writing the paper, we used FlowNet2. In this code, we use RAFT for flow prediction instead. Sorry for the confusion.

dagnichz commented 3 years ago

thanks a lot for your input. I'm trying to use your code to check this method for removing mask areas. In the code example I see you only do object removal. how do you check your code on mask areas like in the paper?

do you just mask each frame and give the mask in the same way or any kind of flag or modification is needed for activation of the code ?

thanks a lot in advance, Nir Dagani

On Mon, Dec 28, 2020 at 4:56 AM Chen Gao notifications@github.com wrote:

RAFT is a recent method (ECCV2020) for optical flow prediction. When writing the paper, we used FlowNet2. In this code, we use RAFT for flow prediction instead. Sorry for the confusion.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/vt-vl-lab/FGVC/issues/28#issuecomment-751556656, or unsubscribe https://github.com/notifications/unsubscribe-auth/AIAYV3G6KDVT4KC755R7XDTSW7XWXANCNFSM4VEKQO7A .

gaochen315 commented 3 years ago

In the real world case, the video is already corrupted. So the input is the corrupted frame and a mask indicating the corrupted region.