Question about ROMP for dataset preprocess

[LarkLeeOnePiece]

It seems the ROMP is useful for static monocular camera and I only predict the "intrinsic" of the camera and set the extrinsic as fix identical Matrix. It works for static monocular camera data. However, the neuman data is the moving camera data. How can we use that for the camera prediction?

[huliangxiao]

Good question!
The neuman data can be considered as if the person is moving while the camera stays still since we only care about the fitting result of the person in the photo.
So it's fine to use ROMP for our task.