Matplotlib: mixed export to pdf and png

Question

I create a figure with matplotlib figure, axes = plt.subplots(nrows=3, ncols=2), plot various stuff axes[0,0].pcolormesh(...) and then export the figure to PDF figure.savefig('figure.pdf') or to PNG figure.savefig('figure.png').

I have to use PNG, because the PDF-file would be huge, but this makes the figure labels and other texts blurry.

Is there a way to export the figure to PDF -- so that labels, etc. are vector graphics -- but with the plots being exported to PNG within the resulting PDF-file? In short: export to PDF, but plots within that PDF to PNG (for small file sizes).

Answer 1

This is one of the huge advantages of Matplotlib over other libraries. If you do:

fig, ax = plt.subplots()
ax.pcolormesh(np.random.randn(500, 500), rasterized=True)
fig.savefig('Test.pdf', dpi=50)

The axes and labels will still be vectors, but the pcolormesh will be rasterized at 50 dpi. Of course for publication you should used a higher dpi, but it still is excellent for reducing large data sets. Note that you will also get aliasing artifacts if you downsample data, so use with caution.