Reading from N5

I have volumes stored in N5 format, and I am serving them via a local http server (using serve). I am also running neuroglancer via npm run dev-server. In neuroglancer, I tried setting the source source to something like n5://http://localhost:5000/$CONTAINER.n5/$DATASET, where $CONTAINER and $DATASET are the names of the n5 container and the dataset I want to view, but this doesn’t work. Specifically, I get Fetching "" resulted in HTTP error 0. How should I get this to work?

[jbms]

The current n5 support is rather limited. It only supports 3d volumes in the format described here: https://github.com/saalfeldlab/n5-viewer

Your data source URL then needs to be of the form:

n5://http://localhost:5000/CONTAINER.n5/DATASET/c0

CONTAINER.n5/DATASET/attributes.json should contain something like:

{ “scales”:[[1,1,1]], “pixelResolution”:{“unit”:“um”,“dimensions”:[0.097,0.097,0.18]} }

CONTAINER.n5/DATASET/c0/attributes.json should not exist (or will be ignored if it does),

and

CONTAINER.n5/DATASET/c0/s0/attributes.json should be a normal n5 dataset attributes.json file (with raw or gzip compression).

The good news is that in a local branch I have already implemented better support for n5 that does not require such a specific directory structure, and I will be pushing it out soon. It is part of a large change that also adds support for 4-d, etc. volumes, but is not quite ready yet.

[jbms]

I just pushed the work in progress to the ndims branch, so you can play with it if you like.

It is still a work in progress, and various things still don’t work — e.g. annotations don’t work at the moment. However, N5 support basically works, though there may be bugs.

It should work with:

• plain single resolution datasets
• multi-resolution datasets using s0, s1, … directories with n5-viewer-style (https://github.com/saalfeldlab/n5-viewer) attributes like “scales”:[[1,1,1],[2,2,1],[4,4,2],[8,8,4],[16,16,9],[32,32,17]], “pixelResolution”:{“unit”:“um”,“dimensions”:[0.097,0.097,0.18]},
• multi-resolution datasets using s0, s1, … directories with bigdataviewer-style (https://github.com/bigdataviewer/bigdataviewer-core/blob/master/BDV N5 format.md) attributes: “downsamplingFactors” : [[1,1,1], [2,2,1], …], “resolution”: [4, 4, 40]

Additionally you can specify attributes: “axes”: [“x”, “y”, “z”, “t”] to specify the axis names, and “units”: [“um”, “um”, “um”, “s”] to specify the units for each dimension.

Neuroglancer now aggregates attributes from all parent directories as well, so the cN directories are no longer necessary or treated specially.