Fix the scale and translation transformations in OME-ZARR metadata #87
Conversation
…anslation based on the voxel resolution
src/main/java/net/preibisch/mvrecon/fiji/spimdata/imgloaders/OMEZarrAttibutes.java
Outdated
Show resolved
Hide resolved
|
@StephanPreibisch I made changes to use the new average calibration. |
| dimensions.get( viewId.getViewSetupId() ), | ||
| //data.getSequenceDescription().getViewDescription( viewId ).getViewSetup().getVoxelSize().dimensionsAsDoubleArray(), | ||
| vx.dimensionsAsDoubleArray(), // resolutionS0 | ||
| vx.unit(), |
There was a problem hiding this comment.
In this case it saves a single viewsetup so I think it is ok to use the voxel attributes from the view setup
|
@cgoina - did you test the fusion results thoroughly? That the fused zarrs display properly in neuroglancer and Fiji? |
|
@StephanPreibisch I only tested for one dataset - the tiff example dataset both in Fiji and neuroglancer. Initially in neuroglancer I got an error because "px" was not valid - so I chose micrometer as default. I have not tried to leave it empty. I will re-test with N5 and TIFF examples to make sure, because it's been a while. |
|
Please try with and without "preserve anisotropy", and with "1x & 2x downsampling", and both combined since it changes the metadata.... thanks so much! |
@StephanPreibisch
This is an attempt to fix the OME-ZARR metadata transformations.
The fix was to also scale the translation with the initial voxel resolution
There's also a fix to use the proper multiscale dataset subpath in case the scale subpath is not numeric, e.g. in case it is 's0', 's1', 's2',...