Channels colors in PSD plot()

Support & Discussions
,
5

I finally tested two solutions :

  1. One directly by modificate a source librairy file and just change colors which are too ligth (and keep MNE colors)

files PATH : …\eenv\lib\site-packages\mne\viz\evoked.py (line 173)

def _rgb(x, y, z):
    """Transform x, y, z values into RGB colors."""
    rgb = np.array([x, y, z]).T
    rgb -= np.nanmin(rgb, 0)
    rgb /= np.maximum(np.nanmax(rgb, 0), 1e-16)  # avoid div by zero
    # Modification to avoid invisible colors :
    # Checking colors which have rgb values near 1 (ligth) for reduce thoses rgb values
    rgb[rgb.sum(axis=1)>2.5] = rgb[rgb.sum(axis=1)>2.5] - 0.3
    return rgb
  1. Like you suggested, by modificate colors with matplotlib, after plot (tested with 4 and 32 channels) :
spect = raw.compute_psd(fmin=0, fmax=50, n_fft=int(raw.info['sfreq']*5))
fig = spect.plot()

channels = raw.ch_names
colors = ['red', 'green', 'blue', 'black'] * int(len(channels)/4)

# Get PSD subplot from figure
fig_psd = fig.axes[0]
# Get Head with sensors subplot from figure
fig_head = fig.axes[1]
# Set sensor color on Head subplot (get the collection)
fig_head.get_children()[0].set_color(colors)
# Set sensor width on Head subplot (get the collection)
fig_head.get_children()[0].set_linewidth(5.0)

# Loop for each channel/color
for nl, (chan, color) in enumerate(zip(channels, colors)):
    # Get the channel line (use nl+2 because of axis)
    line = fig_psd.get_lines()[nl+2]
    # Set color to this line
    line.set_color(color)
    if nl<4:
        # Set color to head lines subplot just for fun
        fig_head.get_lines()[nl].set_color('green')

Thanks for your help.