The existing code assumes that the channel axis is the first axis. This is not
the case for ome-zarr files. In this PR I instead check for the shortest axis
and assume that that contains the channels. This should be safe for most
datasets. (And anyway should soon be deprecated in favour of correct metadata
on input datasets, see #17.)
The existing code assumes that the channel axis is the first axis. This is not the case for ome-zarr files. In this PR I instead check for the shortest axis and assume that that contains the channels. This should be safe for most datasets. (And anyway should soon be deprecated in favour of correct metadata on input datasets, see #17.)