Thanks John
My issue is running a testing pipeline that has a storage limit of 100 MB for each job. Otherwise, we've got hundreds, maybe thousands of deformation fields in static storage for the reason you gave.
My follow up question is, what would happen if I just converted the deformation field to int16, possibly converting it back to float32 in the next step? Would the loss of precision be a big problem when applying the warp? It may be a moot question for my testing use case, since all I care about is that the output is consistent, so consistently imprecise may be good enough. But it may be useful info for others on the list.
Thanks
Paul
P.S. I'm not sure I understand the difference between a displacement field and a deformation field, but I'm sure I can find out.
P.P.S. I still struggle to get backward and forwards, push and pull the right way round, but that's just me: I often get binaries flipped. Aunt Ted and Uncle Sally etc.
|