Viz: evoked.plot_image with the group_by parameter produces faulty sub-panel titles.

[sappelhoff]

Describe the bug

evoked.plot_image with the group_by parameter produces faulty sub-panel titles.

Steps to reproduce

import mne

path = mne.datasets.eegbci.load_data(1, 1)[0]

raw = mne.io.read_raw_edf(path)
mne.datasets.eegbci.standardize(raw)
raw.set_montage("standard_1005")

epochs = mne.make_fixed_length_epochs(raw)
evoked = epochs.average()

sel = mne.channels.make_1020_channel_selections(evoked.info)

fig, axs = plt.subplots(3, 1)
axes = {sel: ax for sel, ax in zip(sel, axs.ravel())}
evoked.plot_image(group_by=sel, axes=axes, show=False)
fig.tight_layout()

Expected results

A plot with 3 sub-panels. the titles should be "Left", "Midline", "Right" ... pretty much as here: https://mne.tools/stable/auto_examples/visualization/roi_erpimage_by_rt.html#sphx-glr-auto-examples-visualization-roi-erpimage-by-rt-py

Actual results

All sub-panels have the title "Left"

This is also the case here: https://mne.tools/stable/auto_tutorials/stats-sensor-space/20_erp_stats.html

Additional information

happens on mne CI in main branch (2022-05-06)

[sappelhoff]

I believe the error resides somewhere here:

mne-python/mne/viz/evoked.py

Lines 238 to 239 in add9e16

	titles = ({channel_type(evoked.info, idx): sel
	for idx in group_by[sel]} if titles is None else titles)

[hoechenberger]

Cool, can you triple-check it's really just the labels that are wrong, or if there's maybe even an issue with the displayed data?

[hoechenberger]

Also for EEG, I don't believe a "channel index" on the Y axis is helping anyone looking at this plot…

[sappelhoff]

Judging from my plots, the data is displayed correctly.

Also for EEG, I don't believe a "channel index" on the Y axis is helping anyone looking at this plot…

you can pass show_names="all" to fix this (see API), but then you have to make the axes bigger as well (or tick label font smaller), else the channel labels overlap by too much