dominant colors

domcolors_no_scikit_learn.py

"""
This Source Code Form is subject to the terms of the Mozilla Public
License, v. 2.0. If a copy of the MPL was not distributed with this
file, You can obtain one at http://mozilla.org/MPL/2.0/.

Copyright (C) 2020 Michael Hall <https://github.com/mikeshardmind>
"""

# Reupload of older stuff in a no longer functional discord bot isolated to single functions
# Summary of edits since isolation from old bot code:
# typing updated
# make this play nice with many ide defaults for unused vars

import numpy as np
import pandas as pd
from PIL import Image
from scipy.cluster.vq import kmeans, whiten


def get_dom_colors(img: Image.Image, count: int) -> list[tuple[int, int, int]]:
    """ Get the top `count` dominant colors from `img` """
    image = np.asarray(img)
    df = pd.DataFrame()
    for idx, c in enumerate(("r", "g", "b")):
        df[c] = pd.Series(image[:,:,idx].flatten())
        df[f"{c}_whiten"] = whiten(df[c])

    cluster_centers, _distortion = kmeans(df[["r_whiten", "g_whiten", "b_whiten"]], count)
    r_std, g_std, b_std = df[["r", "g", "b"]].std()
    return [(int(r * r_std), int(g * g_std), int(b * b_std)) for (r, g, b) in cluster_centers]

domcolors_with_scikit_learn.py

"""
This Source Code Form is subject to the terms of the Mozilla Public
License, v. 2.0. If a copy of the MPL was not distributed with this
file, You can obtain one at http://mozilla.org/MPL/2.0/.
Copyright (C) 2020 Michael Hall <https://github.com/mikeshardmind>
"""

# Reupload of older stuff in a no longer functional discord bot isolated to single functions
# Summary of edits since isolation from old bot code:
# typing updated
# switch to use of scikit-learn for MiniBatchKMeans

import numpy as np
import pandas as pd
from PIL import Image
import scipy.cluster
import sklearn.cluster


def get_dom_colors(img: Image.Image, count: int = 1, iterations: int | None = 1) -> list[tuple[int, int, int]]:
    """
    Get the top `count` dominant colors from `img`
    kmeans will repeat until mo
    """

    image = np.asarray(img)
    df = pd.DataFrame()
    for idx, c in enumerate(("r", "g", "b")):
        df[c] = pd.Series(image[:,:,idx].flatten())
    
    working_matrix = df[["r", "g", "b"]]

    kmeans = sklearn.cluster.MiniBatchKMeans(
        n_init="auto",
        n_clusters=count,
        init="k-means++",
        max_iter=1,
        random_state=0,
    ).fit(working_matrix)
    codes = kmeans.cluster_centers_

    vecs, _dist = scipy.cluster.vq.vq(working_matrix, codes)
    counts, _bins = np.histogram(vecs, len(codes))

    return [tuple(int(code) for code in codes[index]) for index in np.argsort(counts)[::-1]]