trans_gray.py

# encoding: utf-8

import math
from PIL import Image
import random
import  numpy as np
import random

# This is the code of Local Grayscale Transfomation

class LGT(object):

    def __init__(self, probability=0.2, sl=0.02, sh=0.4, r1=0.3):
        self.probability = probability
        self.sl = sl
        self.sh = sh
        self.r1 = r1

    def __call__(self, img):

        new = img.convert("L")   # Convert from here to the corresponding grayscale image
        np_img = np.array(new, dtype=np.uint8)
        img_gray = np.dstack([np_img, np_img, np_img])

        if random.uniform(0, 1) >= self.probability:
            return img

        for attempt in range(100):
            area = img.size[0] * img.size[1]
            target_area = random.uniform(self.sl, self.sh) * area
            aspect_ratio = random.uniform(self.r1, 1 / self.r1)

            h = int(round(math.sqrt(target_area * aspect_ratio)))
            w = int(round(math.sqrt(target_area / aspect_ratio)))

            if w < img.size[1] and h < img.size[0]:
                x1 = random.randint(0, img.size[0] - h)
                y1 = random.randint(0, img.size[1] - w)
                img = np.asarray(img).astype('float')

                img[y1:y1 + h, x1:x1 + w, 0] = img_gray[y1:y1 + h, x1:x1 + w, 0]
                img[y1:y1 + h, x1:x1 + w, 1] = img_gray[y1:y1 + h, x1:x1 + w, 1]
                img[y1:y1 + h, x1:x1 + w, 2] = img_gray[y1:y1 + h, x1:x1 + w, 2]

                img = Image.fromarray(img.astype('uint8'))

                return img

        return img