Is your feature request related to a problem? Please describe.
Currently, to collect whole-slide level predictions and metrics the outputs are written to an intermediate h5 file. This file is not compressed.
Describe the solution you'd like
The current file size of the files can be around 15GB, while likely this level (float32) of precision is not required, and 8-bits with an image compression applied should be sufficient. We could consider using imagecodecs
Is your feature request related to a problem? Please describe. Currently, to collect whole-slide level predictions and metrics the outputs are written to an intermediate h5 file. This file is not compressed.
Describe the solution you'd like The current file size of the files can be around 15GB, while likely this level (float32) of precision is not required, and 8-bits with an image compression applied should be sufficient. We could consider using imagecodecs
Describe alternatives you've considered N/A
Additional context See: https://github.com/NKI-AI/ahcore/blob/6342fe0d14d7444bc9a9e3856ea6339da58b51ea/ahcore/callbacks.py#L265