larsoner@bunk:~/python/MNE-misc-data/seeg$ du -hs
93M	.
larsoner@bunk:~/python/MNE-misc-data/seeg$ git branch
* master
larsoner@bunk:~/python/MNE-misc-data/seeg$ git checkout -b seeg2 alexrockhill/seeg2
Branch 'seeg2' set up to track remote branch 'seeg2' from 'alexrockhill'.
Switched to a new branch 'seeg2'
larsoner@bunk:~/python/MNE-misc-data/seeg$ du -hs
284M	.

This is actually smaller than I expected. Still, it does seem like there are some easy targets for removal that I have never seen used (plus one file with a ~ that is probably safe to remove):

BIN +1.08 MB seeg/sample_seeg/mri/brainmask.mgz~ 
BIN +31.4 MB seeg/sample_seeg/mri/transforms/talairach.m3z 
BIN +5.68 MB seeg/sample_seeg/surf/lh.cvs.tocvs_avg35_inMNI152.sphere.reg
BIN +5.35 MB seeg/sample_seeg/surf/lh.seghead.inflated
BIN +5.68 MB seeg/sample_seeg/surf/lh.white.preaparc
BIN +5.71 MB seeg/sample_seeg/surf/rh.cvs.tocvs_avg35_inMNI152.sphere.reg
BIN +5.71 MB seeg/sample_seeg/surf/rh.white.preaparc 

Removing these brings us down to 225 MB. To me it seems worth it to make the recon "incomplete" to save 20%. WDYT? Any others you think we can safely remove?

I think it's also okay to remove these since if people really want them, they can get them (e.g., talairach.m3z) easily from https://surfer.nmr.mgh.harvard.edu/fswiki/ReconAllTableStableV6.0

If adding another 100 MB here is too much, we could pretty easily split this off into a mne.datasets.seeg.data_path(). Up to you @agramfort ...

If adding another 100 MB here is too much, we could pretty easily split this off into a mne.datasets.seeg.data_path(). Up to you @agramfort ...

I'm fine adding files as needed, it just makes more PRs as we seem to be adding back much of the Freesurfer output. I'll do a minimal version now.

Looks like about 40 MB now just by a rough count scrolling through, I think we could optimize more but this is really nice stuff to have for visualizations

Last request from me is to add back sphere and sphere.reg surfaces. This will be needed for any surface morph operation. With these plus the files you have here I think every built-in MNE function will work.

That extra 20 MB is on you @larsoner ;)

Ok, I think good to merge unless there are any more suggestions

Released 0.15