chromatose

A package for storing and visualizing palettes, and constructing new ones via polynomial interpolation and image extraction. Now pippable!

*This package borrows inspiration from @jmaasch's scatterplots in her package [`sanzo`](https://github.com/jmaasch/sanzo), and uses extraction algorithms from @qTipTip's package [`Pylette`](https://github.com/qTipTip/Pylette).*

`palettes` 🎨


frieda	plath	selah	blonde	honeycombe
crepuscule	leather	gucci	sandbar	fiestaware
chilaquiles	eleven	splenda	gremlin	pollen
medium	juniper	polya	ostrich	fraiche
cheshire	carmine	joan*	pinot	addams
minuit	fugazi	clementine	leda	pugsley
bellhooks	wes	lysergic	pepo	reese
lufte	oolong	pitaya	alice	neko
belle	spiff	yoshi	phoebe	pam
menthol	dwight	riley	mona	eeyore
rory	pudding	marmalade	polaris	trefoil
harmon	abed	shirley	surely	annie
pierce	britta	strogatz	lutz	writhe
perl*	roosh	seafare	heliotrope	moonbow
peanuts	naval	indomie	maggie	rainbow
???	quinone	quinoline	diverging	orb*
warble*	waitomo*	vylette*	pom*	lava*
pumpkin*	pinctada*	nacre*	moxxi*	salvia*
bluefish	BuPu*	holst	joker*	rach*
blink*	betan*	rue*	otterpop*	gummi*
emporium	paired	category20a	category20b	category20c

* available as xxx256

[discrete] [monochrome continuous] [polychromatic continuous] [bokeh]

Viz 👀

ct.palplot( )

Visualizations include swatches, pies, points, lines, scatters. Somewhat helpful for seeing how colors behave on a plot, in dense or scattered visuals. Sometimes colors look great together on a swatch, but not so great in their pointillistic forms.

palette : list or iterable any combination of hex strings or rgb tuples or HTML
bg_color : HTML or hex string background fill color
alpha : fraction between 0.0 and 1.0 alpha transparency of entire palette
shuffle : boolean shuffles palette

ct.palpicker( )

Testing out panel's new colorpicker widget! Useful for adjusting palettes to personal taste.

n : list or integer

Interpolation

ct.palpolate( )

palpolate (pal-(ette inter)-polate): input lists of any size and user can control output size.

There are a multitude of beautiful gradients in packages like bokeh and colorcet that are, for the most part, static. The interpolation scheme here can be used to create entirely new ones given only a few endpoints. This part is still in development, but currently uses linear or polynomial fits in color space metrics RGB, HSL, or HSV. Heatmaps are a good way to visualize the results!

On the left are the input palettes, and on the right are the output palettes all approximately of size 256.

palette : list or iterable any combination of hex strings or rgb tuples or HTML
n_colors : integer approximate desired length of final palette
method : string 'rgb' or 'hsv' or 'hsl' interpolation metric, default 'rgb'
curve : boolean if True, fit t 2ndº polynomial if False, simple linear interpolation
directions : list of 3 strings, 'up' or 'down' each entry corresponds to r, g, b 'up' pushes intermediate values higher (lighter) 'down' pushes intermediate values lower (darker)

Extraction 📷

Extract palette of size n_colors from image given image path using k-means or median cut algorithms.

ct.extract( )

path : string of image path
n_colors : integer number of colors to extract
method : 'kmeans' or 'median' or 'both' algorithm of extraction either k-means clustering or median cut, default k-means
resize : boolean, default True resizing samples a smaller image, speeds up extraction
sort : boolean, default False amateur sort by luminance
show : boolean, default True prints palette and returns panel object

`Color mapping`

map_palette( )

Given palette and 1D quantitative axis, returns list of mapped colors. Uses min and max to first and last of palette

q : 1D quantitative axis, numpy array
palette : color palette to map to
Returns colors: list of mapped colors

map_palette_thresh( )

Given palette and 1D quantitative axis, returns list of mapped colors. Instead of min and max of dataset, uses customized end points.

For example, a dataset may range from (-1, 1), and you want extreme colors to reflect those bounds, not necessarily your dataset's range.

q : 1D quantitative axis, numpy array
palette : color palette to map to
q_min_thresh : minimum threshold, default -infinity
q_max_thresh : maximum threshold, default +infinity
min_color : color to map minimum threshold to (otherwise set to first color in palette)
max_color : color to map maximum threshold to (otherwise set to last color in palette)
nan_color : color to map nan values to
Returns colors: list of mapped colors

map_palette_diverging( )

For customizing diverging palettes with variable breakpoints and thresholds.

For example, a dataset may range from (-x, +y), and you want zero to be the color/point at which the colors diverge.

This function assumes you pass in a balanced palette, but 'midpoint' can approximately adjust this.

q : 1D quantitative axis, numpy array
palette : color palette to map to
q_dividing : dividing numeric value corresponding to divergence in palette
q_min_thresh : minimum threshold, default -infinity
q_max_thresh : maximum threshold, default -infinity
nan_color : color to map nan values to, default black
midpoint : fraction, where to set midpoint of colorbar to, default 0.5
Returns colors: list of mapped colors

Command line tools

Palettes can be retrieved from command line:

>>> chromatose leda
>>> ['#2c2f30', '#8fa7d7', '#afd7d6', '#aa3731', '#f5b3b8']

Palettes can be extracted from command line:

>>> extract egg.png
>>> K-Means:    ['#050002', '#0c1c7b', '#013adc', '#1b143a', '#294254']
>>> Median Cut: ['#040000', '#080006', '#1e1b3c', '#1e2d61', '#0330bd']

Palette extraction from image path using both k-means and median cut
algorithms. Defaults to 5 colors, resized, unsorted, and not displayed.

 path             path to image

 -n , --ncolors   length of palette (int, default 5)
 -r , --resize    resize for efficiency (bool, default True)
 -s , --show      display pillow swatches (bool, default False)