cvdkit

English | 한국어

Colour-vision-deficiency-safe styling for matplotlib: palettes that stay apart under protan, deutan and tritan simulation, markers, dash patterns and hatches cycled with the colours, and a check that finds the series in a figure that only colour tells apart.

Palette and style samples generated by examples/swatches.py. Color is paired with markers, dashes and hatches so the distinctions do not depend on color alone.

Install

cvdkit needs Python 3.12 or newer. Until the first release is on PyPI, install it from a clone of this repository:

uv venv
uv pip install .

uv venv
uv pip install .

Check that it imports:

uv run --no-project python -c "import cvdkit; print(cvdkit.pick(3).colors)"

uv run --no-project python -c "import cvdkit; print(cvdkit.pick(3).colors)"

Use

import matplotlib.pyplot as plt
import cvdkit

p = cvdkit.pick(4)  # 4 colours that stay apart for every simulated deficiency
fig, ax = plt.subplots()
ax.set_prop_cycle(p.cycler())  # colour, marker and dash pattern change together
for k in range(4):
    ax.plot(range(10), [k * x for x in range(10)], label=f"k={k}")
ax.legend()

for c in cvdkit.check(fig):  # pairs that only colour separates, and colour fails
    print(c.a, c.b, c.view, c.color_a, c.color_b)

sim = cvdkit.simulate(fig, "deutan")  # RGB array of the figure as seen with deuteranopia
plt.imsave("figure_deutan.png", sim)

• simulate(obj, kind, severity) takes a figure, an image array or colour specs. kind is "protan", "deutan" or "tritan", and severity runs from 0 to 1, where 1 is dichromacy.
• check(fig) compares the lines, scatter sets and bar sets inside each Axes. A pair is reported when nothing but colour separates them (same marker, dash pattern or hatch) and the colours collide under one of the simulations. Colour-mapped images are not checked; look at them through simulate.
• distinguishable(a, b) and separation(a, b, view) give the same test for two colours.
• pick(n, background) takes the n colours of a palette (Okabe-Ito by default) whose closest pair is farthest apart under every simulation, and drops colours that would be lost against the background. Past what the palette can keep apart, colours repeat and the markers, dashes and hatches separate the series.
• style_cycler(colors) and hatch_cycler(colors) build the cyclers for any colour list.
• palette(name) and colormap(name) give the data below.

When two colours count as distinct

Lines and markers must differ in CIELAB lightness by at least 10, or in hue angle by at least 30 degrees with both colours at chroma 15 or more, under normal vision and under each simulation. Filled areas need a CIEDE2000 difference of at least 10. A difference in saturation alone does not count for lines: #eb6834 and #1baf7a keep a CIEDE2000 distance of about 17 under deuteranopia, more than some pairs in the palettes below, yet both become a mid olive at the same lightness and a thin line cannot show which is which. These thresholds are working values chosen for this package, not results from a study.

Data

docs/swatches.png shows every palette in its original colours, as lines with markers and dashes and as hatched bars (drawn by examples/swatches.py).

Categorical palettes, with the smallest pairwise CIEDE2000 under normal vision and each simulation at severity 1:

name	colours	normal	protan	deutan	tritan	source	terms
okabe_ito	8	21.7	12.2	11.6	10.9	Okabe and Ito; Wong 2011	free to use with credit
tol_bright	7	20.5	14.2	14.8	8.6	Tol	BSD-3-Clause
tol_vibrant	7	18.3	13.7	14.4	8.6	Tol	BSD-3-Clause
tol_muted	9	15.0	11.8	14.3	11.5	Tol	BSD-3-Clause
tol_high_contrast	3	38.9	26.0	25.3	21.2	Tol	BSD-3-Clause
petroff6	6	17.9	10.6	14.7	18.3	Petroff 2021	CC0-1.0
petroff8	8	20.0	9.0	14.5	11.7	Petroff 2021	CC0-1.0
petroff10	10	17.3	11.9	12.6	13.3	Petroff 2021	CC0-1.0

Colormaps: colormap("tol_sunset") and colormap("tol_BuRd") are Tol's diverging maps. For sequential data use matplotlib's cividis (Nuñez et al. 2018) or viridis; Crameri's scientific colour maps (Crameri et al. 2020) are in the cmcrameri package.

The Tol colours are taken from the tol_colors package; its BSD-3-Clause notice is in THIRD_PARTY_NOTICES.md.

References

• Machado, G. M., Oliveira, M. M. and Fernandes, L. A. F. (2009). A physiologically-based model for simulation of color vision deficiency. IEEE Transactions on Visualization and Computer Graphics 15(6), 1291-1298. doi:10.1109/TVCG.2009.113. The simulation matrices are Table 1, from https://www.inf.ufrgs.br/~oliveira/pubs_files/CVD_Simulation/CVD_Simulation.html; the paper does not state the RGB encoding, and cvdkit applies them to linear-light sRGB.
• Brettel, H., Viénot, F. and Mollon, J. D. (1997). Computerized simulation of color appearance for dichromats. Journal of the Optical Society of America A 14(10), 2647. doi:10.1364/JOSAA.14.002647. Background for dichromat simulation; not implemented here.
• Viénot, F., Brettel, H. and Mollon, J. D. (1999). Digital video colourmaps for checking the legibility of displays by dichromats. Color Research & Application 24(4), 243-252. doi:10.1002/(SICI)1520-6378(199908)24:4<243::AID-COL5>3.0.CO;2-3. Background; not implemented.
• Okabe, M. and Ito, K. (2002, revised 2008). Color Universal Design (CUD): Colorblind Barrier Free. https://jfly.uni-koeln.de/color/
• Wong, B. (2011). Points of view: Color blindness. Nature Methods 8(6), 441. doi:10.1038/nmeth.1618.
• Nuñez, J. R., Anderton, C. R. and Renslow, R. S. (2018). Optimizing colormaps with consideration for color vision deficiency to enable accurate interpretation of scientific data. PLOS ONE 13(7), e0199239. doi:10.1371/journal.pone.0199239.
• Crameri, F., Shephard, G. E. and Heron, P. J. (2020). The misuse of colour in science communication. Nature Communications 11, 5444. doi:10.1038/s41467-020-19160-7.
• Petroff, M. A. (2021). Accessible color sequences for data visualization. arXiv:2107.02270. Colour sequences from https://github.com/mpetroff/accessible-color-cycles (CC0-1.0).
• Tol, P. (2021). Colour Schemes. SRON/EPS/TN/09-002, issue 3.2. https://sronpersonalpages.nl/~pault/
• Sharma, G., Wu, W. and Dalal, E. N. (2005). The CIEDE2000 color-difference formula: implementation notes, supplementary test data, and mathematical observations. Color Research & Application 30(1), 21-30. doi:10.1002/col.20070. The tests use its 34 published pairs.

License

MIT. Palette data keeps the terms listed above.