augment.py

#https://nbviewer.jupyter.org/github/aleju/imgaug-doc/blob/master/notebooks/A01%20-%20Load%20and%20Augment%20an%20Image.ipynb

import imageio
from imgaug import augmenters as iaa
import glob

images = []
for filename in glob.glob('data/augment/*.jpg'):
    images += [imageio.imread(filename)]
for filename in glob.glob('data/augment/*.png'):
    images += [imageio.imread(filename)]


seq = iaa.Sequential([
    iaa.Fliplr(0.5), # horizontal flips
    iaa.Crop(percent=(0, 0.1)), # random crops
    # Small gaussian blur with random sigma between 0 and 0.5.
    # But we only blur about 50% of all images.
    iaa.Sometimes(
        0.5,
        iaa.GaussianBlur(sigma=(0, 0.5))
    ),
    # Strengthen or weaken the contrast in each image.
    iaa.LinearContrast((0.75, 1.5)),
    # Add gaussian noise.
    # For 50% of all images, we sample the noise once per pixel.
    # For the other 50% of all images, we sample the noise per pixel AND
    # channel. This can change the color (not only brightness) of the
    # pixels.
    iaa.AdditiveGaussianNoise(loc=0, scale=(0.0, 0.05*255), per_channel=0.5),
    # Make some images brighter and some darker.
    # In 20% of all cases, we sample the multiplier once per channel,
    # which can end up changing the color of the images.
    iaa.Multiply((0.8, 1.2), per_channel=0.2),
    # Apply affine transformations to each image.
    # Scale/zoom them, translate/move them, rotate them and shear them.
    iaa.Affine(
        scale={"x": (0.8, 1.2), "y": (0.8, 1.2)},
        translate_percent={"x": (-0.2, 0.2), "y": (-0.2, 0.2)},
        rotate=(-25, 25),
        shear=(-8, 8)
    )
], random_order=True) # apply augmenters in random order

images_aug = seq(images=images)

c = 0
for image in images_aug:
    # if c > 5:
    #     break
    imageio.imwrite("data/augment/result/" + str(c) + "_2.png", image)
    c += 1