def rotate(images, degree_range=10, n_rotations=1):
"""
Performs a random rotation of a Numpy image tensor.
For more details check:
https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/random_rotation
:param images: (np_array) An array of gray scale images of shape
(num_images, img_width, img_height)
:param degree_range: (int) Rotation range, in degrees.
:param n_rotations: (int) Number of random rotation iterations to perform.
:return: (np_array) Rotated Numpy image tensor.
"""
if len(images.shape) != 3:
raise e.DimensionalityError(
'Invalid shape {} for grayscale images array! Only shapes '
'(num_images, img_width, img_height) are accepted.'.format(
images.shape))
if n_rotations <= 0:
raise ValueError(
'Provide a positive number of rotations. Negative and 0 iterations are not allowed!'
)
rotated = keras_image.random_rotation(images, rg=degree_range)
for i in range(n_rotations - 1):
rotated = np.append(rotated,
keras_image.random_rotation(images,
rg=degree_range),
axis=0)
return rotated
def _augment_image(self, x):
"""
Randomply augment image
"""
assert x.dtype == np.uint8
# common options
co = {
"row_axis": 0,
"col_axis": 1,
"channel_axis": 2,
# can be 'constant', 'nearest', 'reflect', 'wrap'
"fill_mode": "reflect",
"cval": 0.0,
}
o = self._augmentation_options
x = random_rotation(x, o["rotation_max_degrees"], **co)
x = random_shear(x, o["shear_max_degrees"], **co)
x = random_shift(x, o["shift_max_fraction"]["w"],
o["shift_max_fraction"]["h"], **co)
x = random_zoom(x, o["zoom_range"], **co)
x = random_brightness(x, o["brightness_range"])
return x.astype(np.uint8)
def _augment_image(self, x):
"""
Randomply augment image
"""
assert x.dtype == np.uint8
# common options
co = {
"row_axis": 0,
"col_axis": 1,
"channel_axis": 2,
# can be 'constant', 'nearest', 'reflect', 'wrap'
"fill_mode": "nearest",
"cval": 0.0,
}
# default_image_augmenation_options = {
# "rotation_max_degrees": 45,
# "zoom_range": (0.75, 1.25),
# "shift_max_fraction": {"w": 0.25, "h": 0.25},
# "shear_max_degrees": 45,
# "brightness_range": (0.5, 1.5),
# }
o = self.image_augmentation_options
x = random_rotation(x, o["rotation_max_degrees"], **co)
x = random_shear(x, o["shear_max_degrees"], **co)
x = random_shift(x, o["shift_max_fraction"]["w"],
o["shift_max_fraction"]["h"], **co)
x = random_zoom(x, o["zoom_range"], **co)
x = random_brightness(x, o["brightness_range"])
return x
def __call__(self, seed_input):
if tf.is_tensor(seed_input):
seed_input = seed_input.numpy()
seed_input = np.array([
random_rotation(x, self.rg, row_axis=0, col_axis=1, channel_axis=2)
for x in seed_input
])
return tf.constant(seed_input)
def __call__(self, seed_input):
seed_input = [np.asarray(x) for x in seed_input]
seed_input = [
random_rotation(x, self.rg, row_axis=0, col_axis=1, channel_axis=2)
for x in seed_input
]
seed_input = [tf.expand_dims(x, axis=0) for x in seed_input]
return K.concatenate(seed_input, axis=0)
def paint_text(text, w, h, rotate=False, ud=False, multi_fonts=False):
surface = cairo.ImageSurface(cairo.FORMAT_RGB24, w, h)
with cairo.Context(surface) as context:
context.set_source_rgb(1, 1, 1) # White
context.paint()
# this font list works in CentOS 7
if multi_fonts:
fonts = [
'Century Schoolbook', 'Courier', 'STIX', 'URW Chancery L',
'FreeMono'
]
context.select_font_face(
np.random.choice(fonts), cairo.FONT_SLANT_NORMAL,
np.random.choice(
[cairo.FONT_WEIGHT_BOLD, cairo.FONT_WEIGHT_NORMAL]))
else:
context.select_font_face('Courier', cairo.FONT_SLANT_NORMAL,
cairo.FONT_WEIGHT_BOLD)
context.set_font_size(25)
box = context.text_extents(text)
border_w_h = (4, 4)
if box[2] > (w - 2 * border_w_h[1]) or box[3] > (h -
2 * border_w_h[0]):
raise IOError(
('Could not fit string into image.'
'Max char count is too large for given image width.'))
# teach the RNN translational invariance by
# fitting text box randomly on canvas, with some room to rotate
max_shift_x = w - box[2] - border_w_h[0]
max_shift_y = h - box[3] - border_w_h[1]
top_left_x = np.random.randint(0, int(max_shift_x))
if ud:
top_left_y = np.random.randint(0, int(max_shift_y))
else:
top_left_y = h // 2
context.move_to(top_left_x - int(box[0]), top_left_y - int(box[1]))
context.set_source_rgb(0, 0, 0)
context.show_text(text)
buf = surface.get_data()
a = np.frombuffer(buf, np.uint8)
a.shape = (h, w, 4)
a = a[:, :, 0] # grab single channel
a = a.astype(np.float32) / 255
a = np.expand_dims(a, 0)
if rotate:
a = image.random_rotation(a, 3 * (w - top_left_x) / w + 1)
a = speckle(a)
return a
def augmentImages(batch_of_images, batch_of_masks):
for i in range(len(batch_of_images)):
img_and_mask = np.concatenate((batch_of_images[i, ...], batch_of_masks[i,...]), axis=2)
if img_and_mask.ndim == 4: # This assumes single channel data. For multi-channel you'll need
# change this to put all channel in slices channel
orig_shape = img_and_mask.shape
img_and_mask = img_and_mask.reshape((img_and_mask.shape[0:3]))
if np.random.randint(0,10) == 7:
img_and_mask = preprocess.random_rotation(img_and_mask, rg=45, row_axis=0, col_axis=1, channel_axis=2,
fill_mode='constant', cval=0.)
if np.random.randint(0, 5) == 3:
img_and_mask = elastic_transform(img_and_mask, alpha=1000, sigma=80, alpha_affine=50)
if np.random.randint(0, 10) == 7:
img_and_mask = preprocess.random_shift(img_and_mask, wrg=0.2, hrg=0.2, row_axis=0, col_axis=1, channel_axis=2,
fill_mode='constant', cval=0.)
if np.random.randint(0, 10) == 7:
img_and_mask = preprocess.random_shear(img_and_mask, intensity=16, row_axis=0, col_axis=1, channel_axis=2,
fill_mode='constant', cval=0.)
if np.random.randint(0, 10) == 7:
img_and_mask = preprocess.random_zoom(img_and_mask, zoom_range=(0.75, 0.75), row_axis=0, col_axis=1, channel_axis=2,
fill_mode='constant', cval=0.)
if np.random.randint(0, 10) == 7:
img_and_mask = flip_axis(img_and_mask, axis=1)
if np.random.randint(0, 10) == 7:
img_and_mask = flip_axis(img_and_mask, axis=0)
if np.random.randint(0, 10) == 7:
salt_pepper_noise(img_and_mask, salt=0.2, amount=0.04)
if batch_of_images.ndim == 4:
batch_of_images[i, ...] = img_and_mask[...,0:img_and_mask.shape[2]//2]
batch_of_masks[i,...] = img_and_mask[...,img_and_mask.shape[2]//2:]
if batch_of_images.ndim == 5:
img_and_mask = img_and_mask.reshape(orig_shape)
batch_of_images[i, ...] = img_and_mask[...,0:img_and_mask.shape[2]//2, :]
batch_of_masks[i,...] = img_and_mask[...,img_and_mask.shape[2]//2:, :]
# Ensure the masks did not get any non-binary values.
batch_of_masks[batch_of_masks > 0.5] = 1
batch_of_masks[batch_of_masks <= 0.5] = 0
return(batch_of_images, batch_of_masks)
def _augment(image):
image = random_rotation(image,
10,
row_axis=0,
col_axis=1,
channel_axis=2,
fill_mode='constant')
image = random_shift(image,
0.1,
0.1,
row_axis=0,
col_axis=1,
channel_axis=2,
fill_mode='constant')
image = random_brightness(image, (0.9, 1.1))
image = random_zoom(image, (0.85, 1.15),
row_axis=0,
col_axis=1,
channel_axis=2,
fill_mode='constant')
image = tf.image.random_flip_left_right(image)
return image
def rot(img):
return random_rotation(np.array(img),
90,
row_axis=0,
col_axis=1,
channel_axis=2)
import os
import random
import numpy as np
from PIL import Image, ImageStat
from keras.utils import to_categorical
from tensorflow.keras.utils import Sequence
from tensorflow.keras.preprocessing.image import random_brightness, random_shift, random_rotation, random_zoom
# augmentations
brightness_dec = lambda im: random_brightness(im, (0.2, 1))
brightness_inc = lambda im: random_brightness(im, (1, 3))
shift = lambda im: random_shift(
im, wrg=0.2, hrg=0.2, row_axis=0, col_axis=1, channel_axis=2)
rotation = lambda im: random_rotation(
im, 20, row_axis=0, col_axis=1, channel_axis=2)
zoom_in = lambda im: random_zoom(
im, (0.85, 0.85), row_axis=0, col_axis=1, channel_axis=2)
zoom_out = lambda im: random_zoom(
im, (1.15, 1.15), row_axis=0, col_axis=1, channel_axis=2)
class ImageDataGenerator(Sequence):
"""Generates data for Keras
Sequence based data generator. Suitable for building data generator for training and prediction.
"""
def __init__(self,
list_IDs,
sample_k,
mode,
batch_size=32,
dim=(70, 70),
