should `multiscales` be a collection?

[d-v-b]

The purpose of this issue is to discuss whether multiscales should be a collection of multiscale metadata, or just a single instance of multiscale metadata. I.e., should ome-ngff support multiple multiscale images in the same group. My current suspicion is that an overwhelming majority of ome-ngff containers will have just a single element in multiscales, and if that's true, we should re-evaluate whether multiscales should be a JSON array. The argument for not allowing multiple multiscale images to exist in the same multiscale group stems from the complexity it introduces to the spec and implementations. I'm happy to elaborate on this if necessary.

Speaking for the data I manage, I cannot foresee a situation where we would put multiple multiscale images in the same group. But I am curious how many data producers are using or planning to use this feature. Can someone point me to an ome-ngff "in the wild" that has multiple instances of multiscale metadata in multiscales?

[bogovicj]

Chiming in that I don't have a use case for multiple multiscales and would be content for it to be a single instance rather than an array.

[joshmoore]

The only use-case I've ever really proposed is multiple downsamplings, e.g., one for 3D and one for 2D where some clients (like OMERO) can't handle the 3D case currently.

[imagesc-bot]

This issue has been mentioned on Image.sc Forum. There might be relevant details there:

https://forum.image.sc/t/save-a-single-labels-dataset-into-an-ome-zarr/93505/35

[imagesc-bot]

This issue has been mentioned on Image.sc Forum. There might be relevant details there:

https://forum.image.sc/t/whats-the-use-case-for-multiple-multiscales-in-ome-zarr/103267/3