Currently, Zarr/H5 are only considered for inference. We can read test images as Zarr/H5 and create also images in those formats using TEST.BY_CHUNKS variable. The idea is to support training data as well composed by these datatypes, not only multiple images but entire datasets stored in just one file.
Currently,
Zarr
/H5
are only considered for inference. We can read test images asZarr
/H5
and create also images in those formats usingTEST.BY_CHUNKS
variable. The idea is to support training data as well composed by these datatypes, not only multiple images but entire datasets stored in just one file.