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

The shape of flow is inconsistent with the size of the image. #9

Closed Sharpiless closed 1 year ago

Sharpiless commented 1 year ago

For example:

flow_3588.flo in RF-KTest dataset
Timer-x commented 1 year ago

we padded the image of KITTI when trained RAFT , so the shape of flow is same as the image after padding. Use Inputpadder.unpad() in RAFT/core/utils/utils.py to unpad the flow.