Discussion about neuroimaging data packages

Here are some notes from our discussion just now.

The proposal for making it possible to discover data package installations is https://github.com/nipy/nibabel/wiki/data_packages

I put up an issue for that to https://github.com/tryggvib/datapackage . This is the Python interface for using OKFN datapackage.json data packages. Issue : https://github.com/tryggvib/datapackage/issues/26

There are several implementations of 'get data for examples' that I know of:

• https://github.com/nipy/dipy/blob/master/dipy/data/fetcher.py used for example in http://nipy.org/dipy/examples_built/quick_start.html#example-quick-start
• https://github.com/nilearn/nilearn/blob/master/nilearn/datasets.py e.g used in http://nilearn.github.io/building_blocks/manipulating_mr_images.html#datasets

and various others at less advanced stage of development.

• nibabel : https://github.com/nipy/nibabel/blob/master/nibabel/data.py - has a rather ugly previous version of the data package discovery implemented, but that is hardly used.

I think you've already come across http://www.datalad.org/ - a neuroimaging data package tool in development, using git annex - @yarikoptic @hanke - Yarik, Michael - this to put you in touch with Max Ogden and the dat team to see if y'all have insights you can share.

I also mentioned the INCF : http://www.incf.org/ - here I'll ping @jbpoline @chrisfilo about this - JB, Chris - these are notes from a discussion with the dat team about data packages in neuroimaging - I mentioned the INCF and data sharing initiatives.

Also pinging @stefanv

I mentioned the Human Connectome Project : http://www.humanconnectomeproject.org - and the NIfTI data format : http://nifti.nimh.nih.gov/nifti-1

Thanks for the discussion and all the hard thinking y'all are doing.

Read the data_packages wiki proposal and the referenced datapackage issue 26 first. Then compare dipy/data/fetcher.py, nilearn/datasets.py, and nibabel/data.py; the issue does not define a concrete implementation target or acceptance criteria, so what counts as done remains unresolved.