Saving transformations

[rixjilli]

Saving transformations

[jhennies]

Hi!

This is currently not implemented, but it should be feasible to write out all transformations and then run the workflow on another dataset with exactly the same transformations. I can have a look at this in the coming days.

Cheers,
Julian

[rixjilli]

Hi Julian,

That would be really helpful, thanks very much for getting back to me!

Best,
Jillian

[jhennies]

I implemented that AMST saves all transformations when run initially. Then as a second step you can now run these transformations on a second dataset. Only requirement: both datasets need to be the same shape and pixel size.

You find it in the branch pre_align_workflow, it is not added to the master branch yet.

Now you can run the amst_align() function as before (Note: you have to compute your dataset again, the earlier version did not store the required transformations). Afterwards run the apply_amst_transformations() function for the other dataset. Make sure that the "target_folder" is the same folder for both runs (it will create subfolders for the respective results).

For a bit more information have a look at the scripts folder and, in particular, here:
https://github.com/jhennies/amst/blob/pre_align_workflow/scripts/run_amst.py

and here:
https://github.com/jhennies/amst/blob/pre_align_workflow/scripts/run_amst_transforms.py

Let me know if I can help you with anything or if you run into further issues!

[rixjilli]

Great, thanks so much! I look forward to implementing this.