How to export an image with matplotlib

Instead of figure.show() you can use figure.savefig("graph.png") (or any other filename!) to create an image file.

There's a couple things you can add as arguments to .savefig():

• facecolor and edgecolor allow us to set the colour of the main figure or of the edges around the figure. Default to "white".
• bbox_inches allows us to set the size of the figure that we want to save. By default it uses the calculated bounding box of the figure. We can set it to a number, in inches, or most commonly to "tight" to calculate the size of the figure before exporting the image.
• pad_inches allows us to pass in a padding around the figure if bbox_inches is set to "tight". By default it's 0.1.

bbox_inches and pad_inches comparison

Without bbox_inches and pad_inches, the is exported with its calculated bounding box. This makes use of subplots_adjust() and the figsize argument:

figure = plt.figure((figsize=(6,6))
figure.subplots_adjust(bottom=0.1)

...

figure.savefig("graph.png")

You can alternatively set bbox_inches="tight" and pad_inches to a number, and then the bounding box will be calculated before exporting.

The exported image may be bigger than the figsize to accommodate for axes that would normally be clipped:

figure = plt.figure((figsize=(6, 6))
figure.subplots_adjust(bottom=0.1)

...

figure.savefig("graph.png", bbox_inches="tight", pad_inches=2)

Note the pad_inches is quite large. Skipping pad_inches but still setting bbox_inches="tight" will still calculate the bounding box, but use the default padding value which is 0.1:

figure = plt.figure((figsize=(6, 6))
figure.subplots_adjust(bottom=0.1)

...

figure.savefig("graph.png", bbox_inches="tight")

facecolor and edgecolor comparison

facecolor

Produced with figure.savefig("graph.png", facecolor="#5c44fd")

edgecolor

Produced with figure.savefig("graph.png", edgecolor="#5c44fd").

The edgecolor property is the colour of the border. By default figures have no borders, so this won't be visible unless a border is set with figure = plt.figure(linewidth=5):