search

megvii-research / RealFlow

The official implementation of the ECCV 2022 Oral paper: RealFlow: EM-based Realistic Optical Flow Dataset Generation from Videos
Other
93 stars 11 forks source link

Pretrained RAFT checkpoints #6

Closed Sharpiless closed 1 year ago

Sharpiless commented 1 year ago

Wonderful work! But where to download the pretrained C+T model? We only find C and T.

coolbeam commented 1 year ago

Hi, C+T model means that the network is first trianed on C and then finetuned on T. The model RAFT-things is finetuned on T using parameters of RAFT-C. So pretrained C+T model is RAFT-T. Thank you for pointing this out. We will make this clear in the README file.

Sharpiless commented 1 year ago

Got it. Thanks!