megvii-research/RealFlow

How many iterations of GRU used in RAFT forward operation of raft-RFAB.pth pretrained model ?

Closed this issue · 2 comments

I just notice that you set iterations when kitti to 24 and sintel to 32, but have no idea about your RAFT model setting trained by RF-AB dataset.

I just notice that you set iterations when kitti to 24 and sintel to 32, but have no idea about your RAFT model setting trained by RF-AB dataset.

Hi, RAFT model setting trained by RF-AB dataset is the same: 12 when training, 24 when testing on kitti and 32 when testing on sintel. Thank you.

I just notice that you set iterations when kitti to 24 and sintel to 32, but have no idea about your RAFT model setting trained by RF-AB dataset.

Hi, RAFT model setting trained by RF-AB dataset is the same: 12 when training, 24 when testing on kitti and 32 when testing on sintel. Thank you.

Get it. Thanks a lot