Cuboidier [was: transformer]

[JohnTigue]

https://github.com/reconstrue/cuboids/blob/master/by_topic/cuboids/images/cuboid_transformer.jpg

This translates from AIBS RMA RESTful calls for an image stack, to Neuroglancer's static files tree (JSON and binary files which are NumPy array chunks.

This is given a brain-map.org cell_id:

• Will get image stack from brain-map.org
• Will translate to cuboids in neuroglancer precomputed format.

Out of scope:

• No NWB2
• No Exdir

The output of this is a standard preconputed file tree. But that could later be NWB'd or Exdir'd. De-complexes, think unix small specialized tools combined.

• [ ] Check on license

[JohnTigue]

Obviously AllenSDK is involved to read from brain-map.org. Perhaps cloud-volume (Python, BSD2Clause) can be the output.

[JohnTigue]

Use the Xformer code from bossDB. So, it's a Lambda now? Or both Lambda and CLI?

[JohnTigue]

@FrayaMiner has worked out a lot of this. See em_segmentation_flood_filling

[JohnTigue]

There are two things I can think to do with the output of Cuboids Transformer

• Published the file tree that is a precomputed cuboid to S3
• Package the file tree as an NWB:N 2.0 HDF5 file

The former is for consumption by Neuroglancer; the latter is for use in NWB-land.

[JohnTigue]

Decided everything above X-FORM in the above diagrams should be a separate moduele allen_cuboid_cacher, see https://github.com/reconstrue/brightfield/issues/228.

[JohnTigue]

Asked about a license for FreyaMiner's repo.