Matplotlib fonts not found in Google Colab [duplicate] - matplotlib

Question

Using custom fonts in matplotlib locally involves storing the .ttfs in the matplotlib/mpl-data/fonts/ttf/ folder, then calling mpl.font_manager._rebuild(), then setting mpl.rcParams['font.sans-serif'].
Is there any way to do this in Google Colaboratory, where it doesn't seem that this ttf folder is accessible?
For example, I'd like to use the Roboto font. After installing, this would be invoked using mpl.rcParams['font.sans-serif'] = 'Roboto'.

Answer 1

The ttf folder is here:
/usr/local/lib/python3.6/dist-packages/matplotlib/mpl-data/fonts/ttf
So you want to download the ttf there, e.g.:
!wget https://github.com/Phonbopit/sarabun-webfont/raw/master/fonts/thsarabunnew-webfont.ttf -P /usr/local/lib/python3.6/dist-packages/matplotlib/mpl-data/fonts/ttf
matplotlib.font_manager._rebuild()
matplotlib.rc('font', family='TH Sarabun New')
update 2019-12
_rebuild() no longer works. Here's another method which still works.
import matplotlib
import matplotlib.font_manager as fm
!wget https://github.com/Phonbopit/sarabun-webfont/raw/master/fonts/thsarabunnew-webfont.ttf
fm.fontManager.ttflist += fm.createFontList(['thsarabunnew-webfont.ttf'])
matplotlib.rc('font', family='TH Sarabun New')

Answer 2

Wanted to add a full, succinct answer that currently works.
# Download fonts of choice. Here we download Open Sans variants to
# the current directory.
# It's not necessary to download to the share or matplotlib folders:
# /usr/share/fonts/truetype
# /usr/local/lib/python3.6/dist-packages/matplotlib/mpl-data/fonts/ttf
!wget 'https://github.com/google/fonts/raw/master/apache/opensans/OpenSans-Regular.ttf'
!wget 'https://github.com/google/fonts/raw/master/apache/opensans/OpenSans-Light.ttf'
!wget 'https://github.com/google/fonts/raw/master/apache/opensans/OpenSans-SemiBold.ttf'
!wget 'https://github.com/google/fonts/raw/master/apache/opensans/OpenSans-Bold.ttf'
from matplotlib import font_manager as fm, pyplot as plt
# Pick up any fonts in the current directory.
# If you do end up downloading the fonts to /usr/share/fonts/truetype,
# change this to: fm.findSystemFonts()
font_files = fm.findSystemFonts('.')
# Go through and add each to Matplotlib's font cache.
for font_file in font_files:
fm.fontManager.addfont(font_file)
# Use your new font on all your plots.
plt.rc('font', family='Open Sans')
Note, a few times this didn't work properly and the requested font wasn't displayed (even though no error or warning was shown). If that happens, try factory resetting your Colab runtime and running again.

Answer 3

When matplotlib 3.2 is released, it will be easier.
# For now we must upgrade to 3.2 rc first
# !pip install -U --pre matplotlib
import matplotlib as mpl
mpl.font_manager.fontManager.addfont('thsarabunnew-webfont.ttf')
mpl.rc('font', family='TH Sarabun New')

Answer 4

I would like to add my solutions as another reference:
Change the seaborn style
I would strongly recommend this approach as changing the font family can be very troublesome and inconvenient per seaborn design (many of the posts are no longer working on my end in 2022/05). So if you just want to get rid of the stupid default font in matplotlib and seaborn and is OK with Arial, go and type
%matplotlib inline
import matplotlib.style as style
style.use('seaborn-deep')
Changing the font type (borrowed from top answers and tested myself. Restarting the runtime several times if it is not working as expected)
import matplotlib as mpl
import matplotlib.font_manager as fm
from matplotlib import font_manager as fm, pyplot as plt
!wget https://github.com/trishume/OpenTuringCompiler/blob/master/stdlib-sfml/fonts/Times%20New%20Roman.ttf
!wget https://github.com/matomo-org/travis-scripts/blob/master/fonts/Arial.ttf
font_files = fm.findSystemFonts()
# Go through and add each to Matplotlib's font cache.
for font_file in font_files:
fm.fontManager.addfont(font_file)
fm.fontManager.ttflist += fm.createFontList(['Times New Roman.ttf'])
# Use your new font on all your plots.
plt.rc('font', family='serif')
t = np.arange(0.0, 2.0, 0.01)
s = 1 + np.sin(2*np.pi*t)
plt.plot(t, s)
plt.xlabel('time (s)')
plt.ylabel('voltage (mV)')
plt.title('About as simple as it gets, folks')
plt.show()
Updates on 2022/06/08: Method 1 sometimes doesn't work out in Colab, but it works on the local Jupyter Notebook. It seems that explicitly installing and adding the font types is the only way if you want to customize font types on Colab.