Closed manasaV3 closed 1 month ago
Note: at some point a test would be nice
Note: at some point a test would be nice
I agree. I think in addition to this being unit test, this is something we should add to our s3 validation. cc: @daniel-ji
Found the bug:
TomoConverter.scaled_data_transformation
also has to be applied to the base layer of the pyramid here: https://github.com/chanzuckerberg/cryoet-data-portal-backend/blob/66e0ddc0cffdafe8b0d71d678b7443a585337ca2/ingestion_tools/scripts/common/image.py#L273
i.e. change to
pyramid = [self.scaled_data_transformation(self.volume_reader.get_pyramid_base_data())]
We actually had mixed precision pyramids.
Relates to
Description
Fixes the issue of segmentation mask output being a float. The converter selection is determined by if the label is
None
.