Masking annotations in seaborn heatmap

问题

I would like to make a heatmap that has annotation only in specific cells. I though one way to do this would be to make a heatmap with annotations in all cells and then overlay another heatmap that has no annotation but that is masked in the regions that I want the original annotations to be visible:

import numpy as np
import seaborn as sns

par_corr_p = np.array([[1, 2], [3, 4]])
masked_array = np.ma.array(par_corr_p, mask=par_corr_p<2)

fig, ax = plt.subplots()
sns.heatmap(par_corr_p, ax=ax, cmap ='RdBu_r', annot = par_corr_p, center=0, vmin=-5, vmax=5)
sns.heatmap(par_corr_p, mask = masked_array.mask, ax=ax,  cmap ='RdBu_r', center=0, vmin=-5, vmax=5)

However, this is not working - the second heatmap is not covering up the first one:

Please advise

回答1:

I tried a few things, including using numpy.nan or "" in the annot array. Unfortunately they don't work.

This is probably the easiest way. It involves grabbing the texts of the axes, which should only be the labels in annot which sns.heatmap puts there.

import numpy as np
from matplotlib import pyplot as plt
import seaborn as sns

par_corr_p = np.array([[1, 2], [3, 4]])

data = par_corr_p
show_annot_array = data >= 2

fig, ax = plt.subplots()
sns.heatmap(
    ax=ax,
    data=data,
    annot=data,
    cmap ='RdBu_r', center=0, vmin=-5, vmax=5
)
for text, show_annot in zip(ax.texts, (element for row in show_annot_array for element in row)):
    text.set_visible(show_annot)

plt.show()

来源：https://stackoverflow.com/questions/52017735/masking-annotations-in-seaborn-heatmap

标签

python-3.x

matplotlib

seaborn