preprocess.py

import numpy as np
from PIL import Image
from keras.applications.vgg16 import VGG16, preprocess_input, decode_predictions
from keras.preprocessing import image

def binarylab(labels, size, nb_class):
    y = np.zeros((nb_class,size,size))
    for i in range(size):
        for j in range(size):
            y[labels[i][j],i, j] = 1
    return y

def load_data(path, size=224, mode=None):
    img = Image.open(path)
    w,h = img.size
    if w < h:
        if w < size:
            img = img.resize((size, size*h//w))
            w, h = img.size
    else:
        if h < size:
            img = img.resize((size*w//h, size))
            w, h = img.size
    img = img.crop((int((w-size)*0.5), int((h-size)*0.5), int((w+size)*0.5), int((h+size)*0.5)))
    if mode=="original":
        return img

    if mode=="label":
        y = np.array(img, dtype=np.int32)
        mask = y == 255
        y[mask] = 0
        y = binarylab(y, size, 21)
        y = np.expand_dims(y, axis=0)
        return y
    if mode=="data":
        X = image.img_to_array(img)
        X = np.expand_dims(X, axis=0)
        X = preprocess_input(X)
        return X

def generate_arrays_from_file(names, path_to_train, path_to_target, img_size, nb_class):
    while True:
        for name in names:
            Xpath = path_to_train + "{}.jpg".format(name)
            ypath = path_to_target + "{}.png".format(name)
            X = load_data(Xpath, img_size, mode="data")
            y = load_data(ypath, img_size, mode="label")
            yield (X, y)