Get predicted deformation field moving -> fixed about voxelmorph HOT 4 CLOSED

noticed that the prediction result from the registration model is the inverse, i.e. from fixed to moving.

Can you expand on what this mean? The deformation field has to be in the space of the fixed image indeed, because it has to pull the moving image to that space.

perhaps I am misunderstand what you mean, though

We also work with diffeomorphisms in certain models, and computing the inverse field is trivial in that scenario. A bit harder, but still possible. otherwise. Can you tell us exactly which model you are using?

pinging the HyperMorph author @ahoopes , but I know he's quite busy until the CVPR deadline :)

from voxelmorph.

thanks for the quick reply!
I am using the standard hypermorph model and not constraining it to be diffeomorphic.
I want to visualize how the points in the moving image are "pushed" towards the fixed image, this is why i am looking for the inverse.

from voxelmorph.

Okay, I wrote a quick (not well tested) tutorial here: https://colab.research.google.com/drive/1juAJRYhPPDNbO9yRtlc0VGhbIFSuhpJ2?usp=sharing

from voxelmorph.

Ah perfect, thanks a lot!! :)

from voxelmorph.