Convert from Render to neuroglancer precomputed format

[tnguira]

I would like to convert data to the neuroglancer precomputed format from the following pipeline:

raw data -> trakem pipeline of stitching and alignment -> render

Do scripts maybe already exist that convert a render project into the neuroglancer precomputed data format ? If not, do you think that the following would be the way to do it ?

1. Render to disk full resolution slices (or mipmaps if entire 2d slices are too big) using ARGBRenderer.java
2. Convert the rendered 2d slices to precomputed format with scripts similar to the ones used to convert from BigBrain format to precomputed format

(I think that the conversion could also be done directly from Trakem and that the strategy would be the same.)

Thanks a lot for any advice

[trautmane]

Hi,

I'm not aware of any such conversion scripts though I think folks at the Allen Institute have connected neuroglancer to a render web service that dynamically (but likely slowly) renders data.

Your suggested workflow of rendering a stack to disk and then converting it sounds reasonable to me. At Janelia, we render large stacks to disk using the Java BoxClient on our SGE/Univa (now LSF) cluster or the Spark BoxClient on our Spark cluster. Both of these clients ultimately use the ArgbRenderer to render slices to a CATMAID style directory structure that looks like this:

[root directory]/[slice width]x[slice height]/[level]/[z]/[row]/[col].[format]

If you are working with a small enough data set that it all fits into TrakEM2, then presumably you could render directly from there and skip use of the render toolkit components. However, if you have a larger data set or you'd like tile and transform metadata to be easily accessible to other tools, usage of the render toolkit could be helpful.

Let me know if I've misunderstood your goal(s).

Best,
Eric

[tnguira]

Dear Eric,

Thank you very much for these advices. I will follow them.

Best regards,

Tahar

[fcollman]

You should look at this .. it's a c++ program that amongst other things will convert Tiff's to precomputed format.

https://github.com/neurodata/DataManager