How to Save a Seaborn Plot into a File

python

dataframe

matplotlib

best-practices

byAlex Kataev·Feb 20, 2025

To quickly save a Seaborn plot, invoke plt.savefig('filename.ext') right after crafting your masterpiece plot. The 'filename.ext' should carry your desired filename and extension (e.g., .png, .pdf). You won't need any extra imports if you're already working with Seaborn's plot functions. Here's the blueprint:

import seaborn as sns
import matplotlib.pyplot as plt

sns.histplot(data=sns.load_dataset('tips'), x='total_bill')  # Bringing the tips dataset to life
plt.savefig('histogram.png')  # 'Saving Private Histogram'

For paradigms such as FacetGrid or PairGrid, the '.fig' suffix is your wand:

g = sns.PairGrid(sns.load_dataset("penguins"), hue="species")  # Pairing up the penguin species
g.map_diag(sns.histplot)  # Plotting a histogram for the diagonals
g.map_offdiag(sns.scatterplot)  # All other cells get scatter plots
g.add_legend()  # Because "Legends live forever"

g.fig.savefig("penguins_pairplot.png")  # Penguins saved!

Matplotlib back-end: Seaborn's secret sauce

Seaborn produces stellar plots using Matplotlib at its core. Each fancy Seaborn plot is a Matplotlib concoction with an extra layer of aesthetics. When you save a plot with Seaborn, you're essentially calling Matplotlib's savefig() function. Hence, you use plt.savefig() to store your Seaborn plots.

Cumulative wisdom: Scenarios and Caveats

Exploring various scnuarios can be enlightening:

Pixel perfect: Selecting the file format

Tailoring plots to suit their intended usage can bring out their best. Selecting the right file format can be game-changing:

Web: Opt for .png when you don't need to zoom into the plot.
Publications: Use .pdf or .svg for high-quality vector graphics that can be zoomed into without distortion.

Size matters: Configuring figure size

A plot too small or too large can be an eyesore. To resize plots, use the height parameter in pairplot():

sns.pairplot(data=sns.load_dataset("iris"), height=4).savefig("iris_pairplot.pdf")  # A 'tall story' of iris species in PDF

Hue's that? Differentiating by categories

To highlight categories in your plots, the 'hue' parameter is your ally:

sns.scatterplot(data=sns.load_dataset("iris"), x="sepal_length", y="sepal_width", hue="species").get_figure().savefig("iris_scatter_hue.png")  # Saving the 'Rainbow of Iris Species'

Error-proofing: Avoiding common mistakes

Many a tutorial might suggest get_figure() for grid objects like PairGrid or FacetGrid. Such a step leads to an AttributeError, a programming equivalent of a facepalm. Stick to the fig attribute for a smooth coding experience:

# Incorrect
# grid = sns.PairGrid(data)
# grid.get_figure().savefig("should_not_do.png")  # "You shall not pass!"

# Correct
grid = sns.PairGrid(data)
grid.fig.savefig("correct_way.png")  # "There you go!"

Saving extra tips:

Transparency: Use transparent=True to generate plots with no background.
Resolution: Alter the dpi argument for better resolution. Printers appreciate high DPI.
Aspect Ratio: If a plot appears distorted, adjust the aspect parameter when plotting.

Updates: Python and Seaborn

Keep an eye on Python and Seaborn updates - they can influence how savefig operates. It’s a good practice to skim through updated documentation before working with familiar code.

Jupyter: Interactive visualizations

Using Seaborn within a Jupyter notebook? Interact with your plots by exporting them as %matplotlib inline or switch to interactive backends like %matplotlib notebook.