from sklearn.cluster import KMeans
from collections import Counter
import numpy as np
import cv2
from transformers import pipeline, BasePipeline

class ColorExtractionPipeline(BasePipeline):
    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)
        self.image_pipeline = pipeline("image-classification")

    def get_image(self, pil_image):
        nimg = np.array(pil_image)
        image = cv2.cvtColor(nimg, cv2.COLOR_RGB2BGR)
        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
        return image

    def get_labels(self, rimg):
        clf = KMeans(n_clusters=5)
        labels = clf.fit_predict(rimg)
        return labels, clf

    def RGB2HEX(self, color):
        return "#{:02x}{:02x}{:02x}".format(int(color[0]), int(color[1]), int(color[2]))

    def extract_colors(self, pimg):
        img = self.get_image(pimg)
        reshaped_img = img.reshape(img.shape[0] * img.shape[1], img.shape[2])
        labels, clf = self.get_labels(reshaped_img)
        counts = Counter(labels)
        center_colors = clf.cluster_centers_
        ordered_colors = [center_colors[i] for i in counts.keys()]
        hex_colors = [self.RGB2HEX(ordered_colors[i]) for i in counts.keys()]
        closest_color_to_white = min(center_colors, key=lambda c: np.linalg.norm(c - [255, 255, 255]))
        hex_closest_color_to_white = self.RGB2HEX(closest_color_to_white)
        return hex_colors, hex_closest_color_to_white

    def __call__(self, pimg):
        return self.extract_colors(pimg)

color_extraction = ColorExtractionPipeline(task="image-classification")