MRG, ENH: Add ori on MRI slices

[larsoner]

Part of #2995

[hoechenberger]

Code looks good, excited to see how the examples come back :)

[larsoner]

• https://21931-1301584-gh.circle-artifacts.com/0/dev/auto_examples/inverse/plot_mixed_norm_inverse.html#sphx-glr-auto-examples-inverse-plot-mixed-norm-inverse-py
• https://21931-1301584-gh.circle-artifacts.com/0/dev/auto_examples/inverse/plot_time_frequency_mixed_norm_inverse.html#sphx-glr-auto-examples-inverse-plot-time-frequency-mixed-norm-inverse-py
• https://21931-1301584-gh.circle-artifacts.com/0/dev/auto_tutorials/source-modeling/plot_dipole_fit.html#sphx-glr-auto-tutorials-source-modeling-plot-dipole-fit-py+

[agramfort]

this image seems wrong:

https://21931-1301584-gh.circle-artifacts.com/0/dev/_images/sphx_glr_plot_mixed_norm_inverse_003.png

and so is this one to some extent : https://21931-1301584-gh.circle-artifacts.com/0/dev/_images/sphx_glr_plot_dipole_fit_001.png

also for some reason the glyph in glass brain in : https://21931-1301584-gh.circle-artifacts.com/0/dev/auto_examples/inverse/plot_time_frequency_mixed_norm_inverse.html#sphx-glr-auto-examples-inverse-plot-time-frequency-mixed-norm-inverse-py do not show up. I should see 2 sources.

[larsoner]

this image seems wrong:

This is (literally) a matter of perspective (projection), and illustrates why this is so useful. Here is the same image on my machine:

Then rotated to almost the three orthogonal views -- only almost so you can actually tell that there are two arrows...

also for some reason the glyph in glass brain

This is true in master:

https://mne.tools/dev/auto_examples/inverse/plot_time_frequency_mixed_norm_inverse.html#sphx-glr-auto-examples-inverse-plot-time-frequency-mixed-norm-inverse-py

I'll open a separate issue

[agramfort]

indeed. Really perfect !

thx @larsoner

[drammock]

sorry to notice this too late, but there is actually something off about those images. The "stem" that connects the dipole vector to the orthogonal surfaces seems to be partially occluded by the MRI slices. So, hopefully just a z-order problem?

[larsoner]

So, hopefully just a z-order problem?

This is a matplotlib 3D bug. Even when I move them 1 cm inward, the z ordering gets messed up / occluded. I think we have to live with it unless we want to fix their 3D routines...