seaborn/matplotlib custom colormap

Question

I've a 2d numpy array which I want to plot showing different colors to regions (blue for data < 0, green for 0 <= data < 5 and red for data > 5).

In essence, I'm trying to use categorical colors for continuous data based on data range.

Currently I'm using numexpr on data using expression (1 * (data < 0)) + (2 * (data >= 0) & (data < 5)) + (3 * (data >= 5)) . Then using indexed color array/dict ( {1: (0, 0, 255), 2: (0, 255, 0), 3: (255, 0, 0)} ) to compute color values for data. I think this is overkill. There must be an easy way to do this using seaborn/matplot using custom colormaps, which I could not find. Any pointers/sample code would be greatly helpful.

Answer 1

You can create a custom color map and use sns.heatmap with vmin=-1, vmax=6 :

# random data
np.random.seed(1)
a = np.random.uniform(-2,10,(10,10))

from matplotlib import cm, colors as mcolors

# create a custome color map
cmap = mcolors.ListedColormap(['b']+['g']*5 + ['r'], name='abcd', N=7)

# plot heat map, annotation for reference
sns.heatmap(a, annot=True,xticklabels=False, yticklabels=False, cmap=cmap, vmin=-1,vmax=6)

Output: