Codes

# we use keras at first step
from keras.preprocessing.image import ImageDataGenerator, array_to_img, img_to_array, load_img
import os

direction = os.listdir("raw")

for item in direction:
    name = item
    datagen = ImageDataGenerator(
            rotation_range = 40,
            width_shift_range = 0.2,
            height_shift_range = 0.2,
            shear_range = 0.2,
            zoom_range = 0.2,
            horizontal_flip = True,
            fill_mode = "nearest")
# name and direction-----> Merged

    img = load_img(os.path.join('raw', name))
    x = img_to_array(img)
    x = x.reshape((1,) + x.shape)


    i = 0 
    for batch in datagen.flow(x, batch_size = 1, save_to_dir = 'new_augmented', save_prefix = "md", save_format = "jpg") :
        i += 1
        if i > 4 :
            break