in cm/image_datasets.py [0:0]
def random_crop_arr(pil_image, image_size, min_crop_frac=0.8, max_crop_frac=1.0):
min_smaller_dim_size = math.ceil(image_size / max_crop_frac)
max_smaller_dim_size = math.ceil(image_size / min_crop_frac)
smaller_dim_size = random.randrange(min_smaller_dim_size, max_smaller_dim_size + 1)
# We are not on a new enough PIL to support the `reducing_gap`
# argument, which uses BOX downsampling at powers of two first.
# Thus, we do it by hand to improve downsample quality.
while min(*pil_image.size) >= 2 * smaller_dim_size:
pil_image = pil_image.resize(
tuple(x // 2 for x in pil_image.size), resample=Image.BOX
)
scale = smaller_dim_size / min(*pil_image.size)
pil_image = pil_image.resize(
tuple(round(x * scale) for x in pil_image.size), resample=Image.BICUBIC
)
arr = np.array(pil_image)
crop_y = random.randrange(arr.shape[0] - image_size + 1)
crop_x = random.randrange(arr.shape[1] - image_size + 1)
return arr[crop_y : crop_y + image_size, crop_x : crop_x + image_size]