Add .zarr file loading for faster upfront training

[StewartSethA]

Adds a new segment_to_ome.py utility based on existing scroll_to_ome.py for converting segments into .zarr format for faster loading, and an implementation of .zarr array loading in one of the training scripts

[vercel]

The latest updates on your projects. Learn more about Vercel for Git ↗︎

1 Skipped Deployment

Name	Status	Preview	Comments	Updated (UTC)
scrollprize-org	⬜️ Ignored (Inspect)	Visit Preview		Dec 1, 2024 6:33am

[schillij95]

Thank you Seth for your PR

After looking over the scroll2zarr adjustments, i have some questions:

• what is the difference between the scroll_to_ome.py and the segment_to_ome.py?
• scroll_to_ome.py can also take the "input_tiff_dir" argument as path_so_segment/layers as tif/jpg input stack and then output the zarr as the argument "output", ie segment_id.zarr

[graphite-app]

There appears to be a logic error in the zarr_only branch. The function unconditionally returns 1 (error code) regardless of whether tifs2zarr() succeeds or fails. This should be modified to return 0 when err is None and 1 when an error occurs. Consider updating the code to:

if zarr_only:
    print("zarr_only tifs2zarr", tiffdir, zarrdir, chunk_size, slices, maxgb)
    err = tifs2zarr(tiffdir, zarrdir, chunk_size, slices=slices, maxgb=maxgb)
    if err is not None:
        print("error returned:", err)
        return 1
    return 0  # Success case

Suggested change

	if err is not None:
	print("error returned:", err)
	return 1
	if err is not None:
	print("error returned:", err)
	return 1
	return 0 # Success case

Spotted by Diamond

Is this helpful? React 👍 or 👎 to let us know.