mirror of
https://github.com/wassname/mammoviews.git
synced 2026-06-27 16:10:25 +08:00
Dmytro S Lituiev
2208 lines
89 KiB
Python
2208 lines
89 KiB
Python
"""Fairly basic set of tools for real-time data augmentation on image data.
|
|
Can easily be extended to include new transformations,
|
|
new preprocessing methods, etc...
|
|
"""
|
|
from __future__ import absolute_import
|
|
from __future__ import print_function
|
|
|
|
import sys
|
|
import numpy as np
|
|
import re
|
|
from scipy import linalg
|
|
import scipy.ndimage as ndi
|
|
from six.moves import range
|
|
import os
|
|
import threading
|
|
import warnings
|
|
import multiprocessing.pool
|
|
from functools import partial
|
|
from collections import Counter
|
|
|
|
from keras import backend as K
|
|
from keras.utils.data_utils import Sequence
|
|
|
|
from histeq import histeq, ztransform
|
|
try:
|
|
from PIL import ImageEnhance
|
|
from PIL import Image as pil_image
|
|
except ImportError:
|
|
pil_image = None
|
|
|
|
if pil_image is not None:
|
|
_PIL_INTERPOLATION_METHODS = {
|
|
'nearest': pil_image.NEAREST,
|
|
'bilinear': pil_image.BILINEAR,
|
|
'bicubic': pil_image.BICUBIC,
|
|
}
|
|
# These methods were only introduced in version 3.4.0 (2016).
|
|
if hasattr(pil_image, 'HAMMING'):
|
|
_PIL_INTERPOLATION_METHODS['hamming'] = pil_image.HAMMING
|
|
if hasattr(pil_image, 'BOX'):
|
|
_PIL_INTERPOLATION_METHODS['box'] = pil_image.BOX
|
|
# This method is new in version 1.1.3 (2013).
|
|
if hasattr(pil_image, 'LANCZOS'):
|
|
_PIL_INTERPOLATION_METHODS['lanczos'] = pil_image.LANCZOS
|
|
|
|
|
|
import cv2
|
|
|
|
_OPENCV_INTERPOLATION_METHODS = {
|
|
"nearest":cv2.INTER_NEAREST,
|
|
"bilinear":cv2.INTER_LINEAR,
|
|
"linear":cv2.INTER_LINEAR,
|
|
"area":cv2.INTER_AREA,
|
|
"bicubic":cv2.INTER_CUBIC,
|
|
"cubic":cv2.INTER_CUBIC,
|
|
"lanczos":cv2.INTER_LANCZOS4,
|
|
"lanczos4":cv2.INTER_LANCZOS4,
|
|
}
|
|
|
|
import json
|
|
from croppad import crop_pad_center
|
|
from pycocotools.mask import encode, decode
|
|
|
|
|
|
def random_rotation(x, rg, row_axis=1, col_axis=2, channel_axis=0,
|
|
fill_mode='nearest', cval=0.):
|
|
"""Performs a random rotation of a Numpy image tensor.
|
|
|
|
# Arguments
|
|
x: Input tensor. Must be 3D.
|
|
rg: Rotation range, in degrees.
|
|
row_axis: Index of axis for rows in the input tensor.
|
|
col_axis: Index of axis for columns in the input tensor.
|
|
channel_axis: Index of axis for channels in the input tensor.
|
|
fill_mode: Points outside the boundaries of the input
|
|
are filled according to the given mode
|
|
(one of `{'constant', 'nearest', 'reflect', 'wrap'}`).
|
|
cval: Value used for points outside the boundaries
|
|
of the input if `mode='constant'`.
|
|
|
|
# Returns
|
|
Rotated Numpy image tensor.
|
|
"""
|
|
theta = np.pi / 180 * np.random.uniform(-rg, rg)
|
|
rotation_matrix = np.array([[np.cos(theta), -np.sin(theta), 0],
|
|
[np.sin(theta), np.cos(theta), 0],
|
|
[0, 0, 1]])
|
|
|
|
h, w = x.shape[row_axis], x.shape[col_axis]
|
|
transform_matrix = transform_matrix_offset_center(rotation_matrix, h, w)
|
|
x = apply_affine_transform(x, transform_matrix, channel_axis, fill_mode, cval)
|
|
return x
|
|
|
|
|
|
def random_shift(x, wrg, hrg, row_axis=1, col_axis=2, channel_axis=0,
|
|
fill_mode='nearest', cval=0.):
|
|
"""Performs a random spatial shift of a Numpy image tensor.
|
|
|
|
# Arguments
|
|
x: Input tensor. Must be 3D.
|
|
wrg: Width shift range, as a float fraction of the width.
|
|
hrg: Height shift range, as a float fraction of the height.
|
|
row_axis: Index of axis for rows in the input tensor.
|
|
col_axis: Index of axis for columns in the input tensor.
|
|
channel_axis: Index of axis for channels in the input tensor.
|
|
fill_mode: Points outside the boundaries of the input
|
|
are filled according to the given mode
|
|
(one of `{'constant', 'nearest', 'reflect', 'wrap'}`).
|
|
cval: Value used for points outside the boundaries
|
|
of the input if `mode='constant'`.
|
|
|
|
# Returns
|
|
Shifted Numpy image tensor.
|
|
"""
|
|
h, w = x.shape[row_axis], x.shape[col_axis]
|
|
tx = np.random.uniform(-hrg, hrg) * h
|
|
ty = np.random.uniform(-wrg, wrg) * w
|
|
translation_matrix = np.array([[1, 0, tx],
|
|
[0, 1, ty],
|
|
[0, 0, 1]])
|
|
|
|
transform_matrix = translation_matrix # no need to do offset
|
|
x = apply_affine_transform(x, transform_matrix, channel_axis, fill_mode, cval)
|
|
return x
|
|
|
|
|
|
def random_shear(x, intensity, row_axis=1, col_axis=2, channel_axis=0,
|
|
fill_mode='nearest', cval=0.):
|
|
"""Performs a random spatial shear of a Numpy image tensor.
|
|
|
|
# Arguments
|
|
x: Input tensor. Must be 3D.
|
|
intensity: Transformation intensity.
|
|
row_axis: Index of axis for rows in the input tensor.
|
|
col_axis: Index of axis for columns in the input tensor.
|
|
channel_axis: Index of axis for channels in the input tensor.
|
|
fill_mode: Points outside the boundaries of the input
|
|
are filled according to the given mode
|
|
(one of `{'constant', 'nearest', 'reflect', 'wrap'}`).
|
|
cval: Value used for points outside the boundaries
|
|
of the input if `mode='constant'`.
|
|
|
|
# Returns
|
|
Sheared Numpy image tensor.
|
|
"""
|
|
shear = np.random.uniform(-intensity, intensity)
|
|
shear_matrix = np.array([[1, -np.sin(shear), 0],
|
|
[0, np.cos(shear), 0],
|
|
[0, 0, 1]])
|
|
|
|
h, w = x.shape[row_axis], x.shape[col_axis]
|
|
transform_matrix = transform_matrix_offset_center(shear_matrix, h, w)
|
|
x = apply_affine_transform(x, transform_matrix, channel_axis, fill_mode, cval)
|
|
return x
|
|
|
|
|
|
def random_zoom(x, zoom_range, row_axis=1, col_axis=2, channel_axis=0,
|
|
fill_mode='nearest', cval=0.):
|
|
"""Performs a random spatial zoom of a Numpy image tensor.
|
|
|
|
# Arguments
|
|
x: Input tensor. Must be 3D.
|
|
zoom_range: Tuple of floats; zoom range for width and height.
|
|
row_axis: Index of axis for rows in the input tensor.
|
|
col_axis: Index of axis for columns in the input tensor.
|
|
channel_axis: Index of axis for channels in the input tensor.
|
|
fill_mode: Points outside the boundaries of the input
|
|
are filled according to the given mode
|
|
(one of `{'constant', 'nearest', 'reflect', 'wrap'}`).
|
|
cval: Value used for points outside the boundaries
|
|
of the input if `mode='constant'`.
|
|
|
|
# Returns
|
|
Zoomed Numpy image tensor.
|
|
|
|
# Raises
|
|
ValueError: if `zoom_range` isn't a tuple.
|
|
"""
|
|
if len(zoom_range) != 2:
|
|
raise ValueError('`zoom_range` should be a tuple or list of two floats. '
|
|
'Received arg: ', zoom_range)
|
|
|
|
if zoom_range[0] == 1 and zoom_range[1] == 1:
|
|
zx, zy = 1, 1
|
|
else:
|
|
zx, zy = np.random.uniform(zoom_range[0], zoom_range[1], 2)
|
|
zoom_matrix = np.array([[zx, 0, 0],
|
|
[0, zy, 0],
|
|
[0, 0, 1]])
|
|
|
|
h, w = x.shape[row_axis], x.shape[col_axis]
|
|
transform_matrix = transform_matrix_offset_center(zoom_matrix, h, w)
|
|
x = apply_affine_transform(x, transform_matrix, channel_axis, fill_mode, cval)
|
|
return x
|
|
|
|
|
|
def random_channel_shift(x, intensity, channel_axis=0):
|
|
x = np.rollaxis(x, channel_axis, 0)
|
|
min_x, max_x = np.min(x), np.max(x)
|
|
channel_images = [np.clip(x_channel + np.random.uniform(-intensity, intensity), min_x, max_x)
|
|
for x_channel in x]
|
|
x = np.stack(channel_images, axis=0)
|
|
x = np.rollaxis(x, 0, channel_axis + 1)
|
|
return x
|
|
|
|
|
|
def transform_matrix_offset_center(matrix, x, y):
|
|
o_x = float(x) / 2 + 0.5
|
|
o_y = float(y) / 2 + 0.5
|
|
offset_matrix = np.array([[1, 0, o_x], [0, 1, o_y], [0, 0, 1]])
|
|
reset_matrix = np.array([[1, 0, -o_x], [0, 1, -o_y], [0, 0, 1]])
|
|
transform_matrix = np.dot(np.dot(offset_matrix, matrix), reset_matrix)
|
|
return transform_matrix
|
|
|
|
|
|
def apply_affine_transform(x,
|
|
transform_matrix,
|
|
channel_axis=0,
|
|
fill_mode='nearest',
|
|
cval=0.,
|
|
borderMode = cv2.BORDER_TRANSPARENT,
|
|
interp = cv2.INTER_NEAREST,
|
|
use_opencv=False):
|
|
"""Apply the image transformation specified by a matrix.
|
|
|
|
# Arguments
|
|
x: 2D numpy array, single image.
|
|
transform_matrix: Numpy array specifying the geometric transformation.
|
|
channel_axis: Index of axis for channels in the input tensor.
|
|
use_opencv:
|
|
fill_mode: Points outside the boundaries of the input
|
|
are filled according to the given mode
|
|
(one of `{'constant', 'nearest', 'reflect', 'wrap'}`).
|
|
cval: Value used for points outside the boundaries
|
|
of the input if `mode='constant'`.
|
|
borderMode:
|
|
cv2.BORDER_TRANSPARENT
|
|
cv2.BORDER_CONSTANT
|
|
BORDER_REPLICATE
|
|
BORDER_REFLECT
|
|
BORDER_REFLECT101
|
|
BORDER_WRAP
|
|
interp:
|
|
cv2.INTER_NEAREST
|
|
cv2.INTER_LINEAR
|
|
cv2.INTER_AREA
|
|
cv2.INTER_CUBIC
|
|
cv2.INTER_LANCZOS4
|
|
|
|
# Returns
|
|
The transformed version of the input.
|
|
"""
|
|
final_affine_matrix = transform_matrix[:2, :2]
|
|
final_offset = transform_matrix[:2, 2]
|
|
|
|
if use_opencv:
|
|
dsize=x.shape[:2]
|
|
init_shape = x.shape
|
|
#dest = np.ones_like(x) * cval
|
|
x = cv2.warpAffine(x, transform_matrix[:2,:],
|
|
dsize,
|
|
#dest,
|
|
borderValue=cval,
|
|
borderMode = borderMode,
|
|
flags=interp,
|
|
)
|
|
if len(x.shape) < len(init_shape):
|
|
x = x[..., np.newaxis]
|
|
#x = dest
|
|
elif len(x.shape)>2 and channel_axis is not None:
|
|
x = np.rollaxis(x, channel_axis, 0)
|
|
channel_images = [ndi.interpolation.affine_transform(
|
|
x_channel,
|
|
final_affine_matrix,
|
|
final_offset,
|
|
order=0,
|
|
mode=fill_mode,
|
|
cval=cval) for x_channel in x]
|
|
|
|
x = np.stack(channel_images, axis=0)
|
|
x = np.rollaxis(x, 0, channel_axis + 1)
|
|
else:
|
|
x = ndi.interpolation.affine_transform(
|
|
x,
|
|
final_affine_matrix,
|
|
final_offset,
|
|
order=0,
|
|
mode=fill_mode,
|
|
cval=cval)
|
|
return x
|
|
|
|
|
|
def flip_axis(x, axis):
|
|
x = np.asarray(x).swapaxes(axis, 0)
|
|
x = x[::-1, ...]
|
|
x = x.swapaxes(0, axis)
|
|
return x
|
|
|
|
|
|
def array_to_img(x, data_format=None, scale=True):
|
|
"""Converts a 3D Numpy array to a PIL Image instance.
|
|
|
|
# Arguments
|
|
x: Input Numpy array.
|
|
data_format: Image data format.
|
|
scale: Whether to rescale image values
|
|
to be within [0, 255].
|
|
|
|
# Returns
|
|
A PIL Image instance.
|
|
|
|
# Raises
|
|
ImportError: if PIL is not available.
|
|
ValueError: if invalid `x` or `data_format` is passed.
|
|
"""
|
|
if pil_image is None:
|
|
raise ImportError('Could not import PIL.Image. '
|
|
'The use of `array_to_img` requires PIL.')
|
|
x = np.asarray(x, dtype=K.floatx())
|
|
if x.ndim != 3:
|
|
raise ValueError('Expected image array to have rank 3 (single image). '
|
|
'Got array with shape:', x.shape)
|
|
|
|
if data_format is None:
|
|
data_format = K.image_data_format()
|
|
if data_format not in {'channels_first', 'channels_last'}:
|
|
raise ValueError('Invalid data_format:', data_format)
|
|
|
|
# Original Numpy array x has format (height, width, channel)
|
|
# or (channel, height, width)
|
|
# but target PIL image has format (width, height, channel)
|
|
if data_format == 'channels_first':
|
|
x = x.transpose(1, 2, 0)
|
|
if scale:
|
|
x = x + max(-np.min(x), 0)
|
|
x_max = np.max(x)
|
|
if x_max != 0:
|
|
x /= x_max
|
|
x *= 255
|
|
if x.shape[2] == 3:
|
|
# RGB
|
|
return pil_image.fromarray(x.astype('uint8'), 'RGB')
|
|
elif x.shape[2] == 1:
|
|
# grayscale
|
|
return pil_image.fromarray(x[:, :, 0].astype('uint8'), 'L')
|
|
else:
|
|
raise ValueError('Unsupported channel number: ', x.shape[2])
|
|
|
|
|
|
def img_to_array(img, data_format=None):
|
|
"""Converts a PIL Image instance to a Numpy array.
|
|
|
|
# Arguments
|
|
img: PIL Image instance.
|
|
data_format: Image data format.
|
|
|
|
# Returns
|
|
A 3D Numpy array.
|
|
|
|
# Raises
|
|
ValueError: if invalid `img` or `data_format` is passed.
|
|
"""
|
|
if data_format is None:
|
|
data_format = K.image_data_format()
|
|
if data_format not in {'channels_first', 'channels_last'}:
|
|
raise ValueError('Unknown data_format: ', data_format)
|
|
# Numpy array x has format (height, width, channel)
|
|
# or (channel, height, width)
|
|
# but original PIL image has format (width, height, channel)
|
|
x = np.asarray(img, dtype=K.floatx())
|
|
if len(x.shape) == 3:
|
|
if data_format == 'channels_first':
|
|
x = x.transpose(2, 0, 1)
|
|
elif len(x.shape) == 2:
|
|
if data_format == 'channels_first':
|
|
x = x.reshape((1, x.shape[0], x.shape[1]))
|
|
else:
|
|
x = x.reshape((x.shape[0], x.shape[1], 1))
|
|
else:
|
|
raise ValueError('Unsupported image shape: ', x.shape)
|
|
return x
|
|
|
|
|
|
def load_img(path, grayscale=False, color_mode='rgb', target_size=None,
|
|
interpolation='nearest', driver='opencv'):
|
|
"""Loads an image using a defined module
|
|
# Arguments
|
|
path: Path to image file.
|
|
color_mode: One of "grayscale", "rbg", "rgba", "bgr", "bgra". Default: "rgb".
|
|
The desired image format.
|
|
target_size: Either `None` (default to original size)
|
|
or tuple of ints `(img_height, img_width)`.
|
|
interpolation: Interpolation method used to resample the image if the
|
|
target size is different from that of the loaded image.
|
|
Supported methods are "nearest", "bilinear", and "bicubic".
|
|
Driver-dependent options are "lanczos", "area", "box" and
|
|
"hamming". For details, refer to documentation of `load_img_pil()`
|
|
and `load_img_opencv()`. By default, "nearest" is used.
|
|
"""
|
|
if driver.lower() in ('cv2','opencv'):
|
|
img = load_img_opencv(path, grayscale=grayscale, color_mode=color_mode,
|
|
target_size=target_size, interpolation=interpolation)
|
|
elif driver.lower() in ('pil', 'pillow'):
|
|
img = load_img_pil(path, grayscale=grayscale, color_mode=color_mode,
|
|
target_size=target_size, interpolation=interpolation)
|
|
return img
|
|
|
|
|
|
def load_img_opencv(path, grayscale=False, color_mode='rgb', target_size=None,
|
|
interpolation='nearest'):
|
|
"""Loads an image using opencv format.
|
|
# Arguments
|
|
path: Path to image file.
|
|
color_mode: One of "grayscale", "rbg", "rgba", "bgr", "bgra". Default: "rgb".
|
|
The desired image format.
|
|
target_size: Either `None` (default to original size)
|
|
or tuple of ints `(img_height, img_width)`.
|
|
interpolation: Interpolation method used to resample the image if the
|
|
target size is different from that of the loaded image.
|
|
Supported methods are "nearest", "bilinear", "bicubic", "lanczos", "area".
|
|
By default, "nearest" is used.
|
|
# Returns
|
|
A numpy array instance.
|
|
# Raises
|
|
ImportError: if PIL is not available.
|
|
ValueError: if interpolation method is not supported.
|
|
"""
|
|
if grayscale is True:
|
|
warnings.warn('grayscale is deprecated. Please use '
|
|
'color_mode = "grayscale"')
|
|
color_mode = 'grayscale'
|
|
|
|
img = cv2.imread(path, cv2.IMREAD_ANYDEPTH)
|
|
|
|
if len(img.shape)==2:
|
|
img_mode = 'grayscale'
|
|
elif len(img.shape) == 3:
|
|
if img.shape[-1]==3:
|
|
img_mode = 'bgr'
|
|
elif img.shape[-1]==4:
|
|
img_mode = 'bgra'
|
|
|
|
if img_mode != color_mode:
|
|
if color_mode.startswith('gray'):
|
|
if img_mode == 'bgr':
|
|
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
|
|
elif img_mode == 'bgra':
|
|
img = cv2.cvtColor(img, cv2.COLOR_BGRA2GRAY)
|
|
elif color_mode == 'rgba':
|
|
if img_mode == 'bgr':
|
|
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGBA)
|
|
elif img_mode == 'grayscale':
|
|
img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGBA)
|
|
elif color_mode == 'rgb':
|
|
if img_mode == 'bgr':
|
|
img = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)
|
|
elif img_mode == 'bgra':
|
|
img = cv2.cvtColor(img, cv2.COLOR_BGRA2RGB)
|
|
elif img_mode == 'grayscale':
|
|
img = cv2.cvtColor(img, cv2.COLOR_GRAY2RGB)
|
|
else:
|
|
raise ValueError('color_mode must be "grayscale", "rbg", or "rgba"')
|
|
|
|
if target_size is not None:
|
|
width_height_tuple = (target_size[1], target_size[0])
|
|
if img.shape[:2][::-1] != width_height_tuple:
|
|
if interpolation not in _PIL_INTERPOLATION_METHODS:
|
|
raise ValueError(
|
|
'Invalid interpolation method {} specified. Supported '
|
|
'methods are {}'.format(
|
|
interpolation,
|
|
", ".join(_OPENCV_INTERPOLATION_METHODS.keys())))
|
|
resample = _OPENCV_INTERPOLATION_METHODS[interpolation]
|
|
img = cv2.resize(img, width_height_tuple, resample)
|
|
return img
|
|
|
|
|
|
def load_img_pil(path, grayscale=False, color_mode='rgb', target_size=None,
|
|
interpolation='nearest'):
|
|
"""Loads an image into PIL format.
|
|
# Arguments
|
|
path: Path to image file.
|
|
color_mode: One of "grayscale", "rbg", "rgba". Default: "rgb".
|
|
The desired image format.
|
|
target_size: Either `None` (default to original size)
|
|
or tuple of ints `(img_height, img_width)`.
|
|
interpolation: Interpolation method used to resample the image if the
|
|
target size is different from that of the loaded image.
|
|
Supported methods are "nearest", "bilinear", and "bicubic".
|
|
If PIL version 1.1.3 or newer is installed, "lanczos" is also
|
|
supported. If PIL version 3.4.0 or newer is installed, "box" and
|
|
"hamming" are also supported. By default, "nearest" is used.
|
|
# Returns
|
|
A PIL Image instance.
|
|
# Raises
|
|
ImportError: if PIL is not available.
|
|
ValueError: if interpolation method is not supported.
|
|
"""
|
|
if grayscale is True:
|
|
warnings.warn('grayscale is deprecated. Please use '
|
|
'color_mode = "grayscale"')
|
|
color_mode = 'grayscale'
|
|
if pil_image is None:
|
|
raise ImportError('Could not import PIL.Image. '
|
|
'The use of `array_to_img` requires PIL.')
|
|
img = pil_image.open(path)
|
|
|
|
if color_mode == 'grayscale':
|
|
if img.mode not in ('L', 'I;16'):
|
|
img = img.convert('L')
|
|
elif color_mode == 'rgba':
|
|
if img.mode != 'RGBA':
|
|
img = img.convert('RGBA')
|
|
elif color_mode == 'rgb':
|
|
if img.mode != 'RGB':
|
|
if img.mode not in ('I;16'):
|
|
img = img.convert('RGB')
|
|
else:
|
|
img = np.asarray(img)
|
|
img = img * (255.0/ max(1.0, img.max()))
|
|
img = np.stack([img.astype('uint8')]*3, axis=-1)
|
|
img = array_to_img(img)
|
|
else:
|
|
raise ValueError('color_mode must be "grayscale", "rbg", or "rgba"')
|
|
if target_size is not None:
|
|
width_height_tuple = (target_size[1], target_size[0])
|
|
if img.size != width_height_tuple:
|
|
if interpolation not in _PIL_INTERPOLATION_METHODS:
|
|
raise ValueError(
|
|
'Invalid interpolation method {} specified. Supported '
|
|
'methods are {}'.format(
|
|
interpolation,
|
|
", ".join(_PIL_INTERPOLATION_METHODS.keys())))
|
|
resample = _PIL_INTERPOLATION_METHODS[interpolation]
|
|
img = img.resize(width_height_tuple, resample)
|
|
return img
|
|
|
|
|
|
def array_to_img(x, data_format=None, scale=True):
|
|
"""Converts a 3D Numpy array to a PIL Image instance.
|
|
# Arguments
|
|
x: Input Numpy array.
|
|
data_format: Image data format.
|
|
either "channels_first" or "channels_last".
|
|
scale: Whether to rescale image values
|
|
to be within `[0, 255]`.
|
|
# Returns
|
|
A PIL Image instance.
|
|
# Raises
|
|
ImportError: if PIL is not available.
|
|
ValueError: if invalid `x` or `data_format` is passed.
|
|
"""
|
|
if pil_image is None:
|
|
raise ImportError('Could not import PIL.Image. '
|
|
'The use of `array_to_img` requires PIL.')
|
|
x = np.asarray(x, dtype=K.floatx())
|
|
if x.ndim != 3:
|
|
raise ValueError('Expected image array to have rank 3 (single image). '
|
|
'Got array with shape:', x.shape)
|
|
|
|
if data_format is None:
|
|
data_format = K.image_data_format()
|
|
if data_format not in {'channels_first', 'channels_last'}:
|
|
raise ValueError('Invalid data_format:', data_format)
|
|
|
|
# Original Numpy array x has format (height, width, channel)
|
|
# or (channel, height, width)
|
|
# but target PIL image has format (width, height, channel)
|
|
if data_format == 'channels_first':
|
|
x = x.transpose(1, 2, 0)
|
|
if scale:
|
|
x = x + max(-np.min(x), 0)
|
|
x_max = np.max(x)
|
|
if x_max != 0:
|
|
x /= x_max
|
|
x *= 255
|
|
if x.shape[2] == 4:
|
|
# RGBA
|
|
return pil_image.fromarray(x.astype('uint8'), 'RGBA')
|
|
elif x.shape[2] == 3:
|
|
# RGB
|
|
return pil_image.fromarray(x.astype('uint8'), 'RGB')
|
|
elif x.shape[2] == 1:
|
|
# grayscale
|
|
if np.all(x < 255):
|
|
return pil_image.fromarray(x[:, :, 0].astype('uint8'), 'L')
|
|
else:
|
|
return pil_image.fromarray(x[:, :, 0].astype('uint16'), 'I;16')
|
|
else:
|
|
raise ValueError('Unsupported channel number: ', x.shape[2])
|
|
|
|
|
|
def list_pictures(directory, ext='jpg|jpeg|bmp|png|ppm'):
|
|
return [os.path.join(root, f)
|
|
for root, _, files in os.walk(directory) for f in files
|
|
if re.match(r'([\w]+\.(?:' + ext + '))', f)]
|
|
|
|
|
|
class ImageDataGenerator(object):
|
|
"""Generate minibatches of image data with real-time data augmentation.
|
|
|
|
# Arguments
|
|
featurewise_center: set input mean to 0 over the dataset.
|
|
samplewise_center: set each sample mean to 0.
|
|
featurewise_std_normalization: divide inputs by std of the dataset.
|
|
samplewise_std_normalization: divide each input by its std.
|
|
zca_whitening: apply ZCA whitening.
|
|
zca_epsilon: epsilon for ZCA whitening. Default is 1e-6.
|
|
rotation_range: degrees (0 to 180).
|
|
width_shift_range: fraction of total width.
|
|
height_shift_range: fraction of total height.
|
|
shear_range: shear intensity (shear angle in radians).
|
|
zoom_range: amount of zoom. if scalar z, zoom will be randomly picked
|
|
in the range [1-z, 1+z]. A sequence of two can be passed instead
|
|
to select this range.
|
|
channel_shift_range: shift range for each channel.
|
|
fill_mode: points outside the boundaries are filled according to the
|
|
given mode ('constant', 'nearest', 'reflect' or 'wrap'). Default
|
|
is 'nearest'.
|
|
cval: value used for points outside the boundaries when fill_mode is
|
|
'constant'. Default is 0.
|
|
horizontal_flip: whether to randomly flip images horizontally.
|
|
vertical_flip: whether to randomly flip images vertically.
|
|
rescale: rescaling factor. If None or 0, no rescaling is applied,
|
|
otherwise we multiply the data by the value provided. This is
|
|
applied after the `preprocessing_function` (if any provided)
|
|
but before any other transformation.
|
|
preprocessing_function: function that will be implied on each input.
|
|
The function will run before any other modification on it.
|
|
The function should take one argument:
|
|
one image (Numpy tensor with rank 3),
|
|
and should output a Numpy tensor with the same shape.
|
|
data_format: 'channels_first' or 'channels_last'. In 'channels_first' mode, the channels dimension
|
|
(the depth) is at index 1, in 'channels_last' mode it is at index 3.
|
|
It defaults to the `image_data_format` value found in your
|
|
Keras config file at `~/.keras/keras.json`.
|
|
If you never set it, then it will be "channels_last".
|
|
"""
|
|
|
|
def __init__(self,
|
|
featurewise_center=False,
|
|
samplewise_center=False,
|
|
featurewise_std_normalization=False,
|
|
samplewise_std_normalization=False,
|
|
zca_whitening=False,
|
|
zca_epsilon=1e-6,
|
|
rotation_range=0.,
|
|
width_shift_range=0.,
|
|
height_shift_range=0.,
|
|
shear_range=0.,
|
|
zoom_range=0.,
|
|
channel_shift_range=0.,
|
|
fill_mode='nearest',
|
|
cval=0.,
|
|
horizontal_flip=False,
|
|
vertical_flip=False,
|
|
rescale=None,
|
|
preprocessing_function=None,
|
|
postprocessing_function=None,
|
|
histeq_alpha = False,
|
|
contrast_exp = 1.2,
|
|
contrast = None,
|
|
truncate_quantile = None,
|
|
z_transform = None,
|
|
#noise=None
|
|
data_format=None):
|
|
if data_format is None:
|
|
data_format = K.image_data_format()
|
|
self.featurewise_center = featurewise_center
|
|
self.samplewise_center = samplewise_center
|
|
self.featurewise_std_normalization = featurewise_std_normalization
|
|
self.samplewise_std_normalization = samplewise_std_normalization
|
|
self.zca_whitening = zca_whitening
|
|
self.zca_epsilon = zca_epsilon
|
|
self.rotation_range = rotation_range
|
|
self.width_shift_range = width_shift_range
|
|
self.height_shift_range = height_shift_range
|
|
self.shear_range = shear_range
|
|
self.zoom_range = zoom_range
|
|
self.channel_shift_range = channel_shift_range
|
|
self.fill_mode = fill_mode
|
|
self.cval = cval
|
|
self.horizontal_flip = horizontal_flip
|
|
self.vertical_flip = vertical_flip
|
|
self.rescale = rescale
|
|
|
|
self.histeq_alpha = histeq_alpha
|
|
self.contrast = contrast
|
|
self.contrast_exp = contrast_exp
|
|
self.ztransform = z_transform
|
|
self.truncate_quantile = truncate_quantile
|
|
|
|
#self.noise_specs = {} if noise is None else noise
|
|
#self.intensity_gain = intensity_gain
|
|
|
|
self.preprocessing_function = preprocessing_function
|
|
self.postprocessing_function = postprocessing_function
|
|
|
|
if data_format not in {'channels_last', 'channels_first'}:
|
|
raise ValueError('`data_format` should be `"channels_last"` (channel after row and '
|
|
'column) or `"channels_first"` (channel before row and column). '
|
|
'Received arg: ', data_format)
|
|
self.data_format = data_format
|
|
if data_format == 'channels_first':
|
|
self.channel_axis = 1
|
|
self.row_axis = 2
|
|
self.col_axis = 3
|
|
if data_format == 'channels_last':
|
|
self.channel_axis = 3
|
|
self.row_axis = 1
|
|
self.col_axis = 2
|
|
|
|
self.mean = None
|
|
self.std = None
|
|
self.principal_components = None
|
|
|
|
if np.isscalar(zoom_range):
|
|
self.zoom_range = [1 - zoom_range, 1 + zoom_range]
|
|
elif len(zoom_range) == 2:
|
|
self.zoom_range = [zoom_range[0], zoom_range[1]]
|
|
else:
|
|
raise ValueError('`zoom_range` should be a float or '
|
|
'a tuple or list of two floats. '
|
|
'Received arg: ', zoom_range)
|
|
|
|
def flow(self, x, y=None, batch_size=32, shuffle=True, seed=None,
|
|
save_to_dir=None, save_prefix='', save_format='png', color_mode='rgb'):
|
|
return NumpyArrayIterator(
|
|
x, y, self,
|
|
batch_size=batch_size,
|
|
shuffle=shuffle,
|
|
seed=seed,
|
|
data_format=self.data_format,
|
|
save_to_dir=save_to_dir,
|
|
save_prefix=save_prefix,
|
|
save_format=save_format,
|
|
postprocessing_function=self.postprocessing_function,
|
|
color_mode=color_mode)
|
|
|
|
def flow_from_directory(self, directory,
|
|
target_size=(256, 256), color_mode='rgb',
|
|
classes=None, class_mode='categorical',
|
|
batch_size=32, shuffle=True, seed=None,
|
|
save_to_dir=None,
|
|
save_prefix='',
|
|
save_format='png',
|
|
follow_links=False,
|
|
stratify=None,
|
|
oversampling=False,
|
|
subsample_factor= None,
|
|
subsample_num = None,
|
|
):
|
|
return DirectoryIterator(
|
|
directory, self,
|
|
target_size=target_size, color_mode=color_mode,
|
|
classes=classes, class_mode=class_mode,
|
|
data_format=self.data_format,
|
|
batch_size=batch_size, shuffle=shuffle, seed=seed,
|
|
save_to_dir=save_to_dir,
|
|
save_prefix=save_prefix,
|
|
save_format=save_format,
|
|
postprocessing_function=self.postprocessing_function,
|
|
follow_links=follow_links,
|
|
stratify=stratify,
|
|
oversampling=oversampling,
|
|
subsample_factor=subsample_factor,
|
|
subsample_num=subsample_num,
|
|
)
|
|
|
|
|
|
def flow_patches(self, fn_img, fn_pnt,
|
|
point_sampler,
|
|
target_size=(256, 256),
|
|
color_mode='grayscale',
|
|
batch_size=4,
|
|
patches_per_image=1,
|
|
shuffle=True,
|
|
seed=0,
|
|
dtype='uint16',
|
|
fill_mode = 'reflect',
|
|
label_freq={1:5, 2:10},
|
|
postprocessing_functions = [None, None],
|
|
output_indices=False,
|
|
):
|
|
|
|
return PatchIterator(fn_img, fn_pnt,
|
|
point_sampler,
|
|
image_data_generator = self,
|
|
batch_size=batch_size,
|
|
patches_per_image=patches_per_image,
|
|
shuffle=shuffle,
|
|
color_mode=color_mode,
|
|
seed=seed,
|
|
patch_size = target_size,
|
|
dtype=dtype,
|
|
mode = self.fill_mode if fill_mode is None else fill_mode,
|
|
label_freq=label_freq,
|
|
augmentation=True,
|
|
postprocessing_functions=postprocessing_functions,
|
|
output_indices=output_indices,
|
|
)
|
|
|
|
def flow_memmap(self,
|
|
root_dir, csv_file,
|
|
classes = ["Control", "Case",],
|
|
binary=True,
|
|
postprocessing_functions=None,
|
|
nsamples = None,
|
|
batch_size = 1,
|
|
shuffle = False,
|
|
seed = None,
|
|
postprocessing_function=None,
|
|
stratify=None,
|
|
oversampling=False,
|
|
subsample_factor=None,
|
|
subsample_num=None,
|
|
batch_rate=1,
|
|
dtype = K.floatx(),
|
|
color_mode=None,
|
|
label_col = "label",
|
|
filename_col = "filename",
|
|
encode_label=None,
|
|
):
|
|
return MemMapIterator(root_dir, csv_file,
|
|
classes = classes,
|
|
image_data_generator = self,
|
|
data_format = self.data_format,
|
|
binary=binary,
|
|
transform=postprocessing_functions,
|
|
nsamples = None,
|
|
batch_size = batch_size,
|
|
shuffle = shuffle,
|
|
seed = seed,
|
|
stratify=stratify,
|
|
oversampling=oversampling,
|
|
subsample_factor=subsample_factor,
|
|
subsample_num=subsample_num,
|
|
batch_rate=batch_rate,
|
|
dtype = dtype,
|
|
color_mode=color_mode,
|
|
label_col = label_col,
|
|
filename_col = filename_col,
|
|
encode_label=encode_label,
|
|
)
|
|
def standardize(self, x):
|
|
"""Apply the normalization configuration to a batch of inputs.
|
|
|
|
# Arguments
|
|
x: batch of inputs to be normalized.
|
|
|
|
# Returns
|
|
The inputs, normalized.
|
|
"""
|
|
if self.histeq_alpha:
|
|
alpha=np.random.random()
|
|
if hasattr(self.histeq_alpha, '__len__'):
|
|
#print("self.histeq_alpha,", self.histeq_alpha,)
|
|
for ii,aa in enumerate(self.histeq_alpha):
|
|
if aa:
|
|
x[:,:,ii] = histeq(x[:,:,ii], bitdepth=16, mask=None, alpha=alpha)
|
|
else:
|
|
x = histeq(x, bitdepth=16, mask=None, alpha=alpha)
|
|
if self.ztransform:
|
|
mask = x>0
|
|
mask &= x<x.max()
|
|
if self.contrast is not None:
|
|
contrast = np.clip(self.contrast_exp ** np.random.normal(0,1),
|
|
*self.contrast)
|
|
else:
|
|
contrast = None
|
|
x = ztransform(x, mask=None, contrast=contrast,
|
|
truncate_quantile=self.truncate_quantile)
|
|
if self.preprocessing_function:
|
|
x = self.preprocessing_function(x)
|
|
if self.rescale is not None:
|
|
if hasattr(self.rescale, '__len__'):
|
|
for ii, rr in enumerate(self.rescale):
|
|
if rr is not None:
|
|
x[:,:,ii] = rr*x[:,:,ii]
|
|
else:
|
|
#print("x", x.dtype)
|
|
#print("self.rescale", type(self.rescale), self.rescale)
|
|
x *= self.rescale
|
|
# x is a single image, so it doesn't have image number at index 0
|
|
img_channel_axis = self.channel_axis - 1
|
|
if self.samplewise_center:
|
|
x -= np.mean(x, axis=img_channel_axis, keepdims=True)
|
|
if self.samplewise_std_normalization:
|
|
x /= (np.std(x, axis=img_channel_axis, keepdims=True) + 1e-7)
|
|
|
|
if self.featurewise_center:
|
|
if self.mean is not None:
|
|
x -= self.mean
|
|
else:
|
|
warnings.warn('This ImageDataGenerator specifies '
|
|
'`featurewise_center`, but it hasn\'t'
|
|
'been fit on any training data. Fit it '
|
|
'first by calling `.fit(numpy_data)`.')
|
|
if self.featurewise_std_normalization:
|
|
if self.std is not None:
|
|
x /= (self.std + 1e-7)
|
|
else:
|
|
warnings.warn('This ImageDataGenerator specifies '
|
|
'`featurewise_std_normalization`, but it hasn\'t'
|
|
'been fit on any training data. Fit it '
|
|
'first by calling `.fit(numpy_data)`.')
|
|
if self.zca_whitening:
|
|
if self.principal_components is not None:
|
|
flatx = np.reshape(x, (-1, np.prod(x.shape[-3:])))
|
|
whitex = np.dot(flatx, self.principal_components)
|
|
x = np.reshape(whitex, x.shape)
|
|
else:
|
|
warnings.warn('This ImageDataGenerator specifies '
|
|
'`zca_whitening`, but it hasn\'t'
|
|
'been fit on any training data. Fit it '
|
|
'first by calling `.fit(numpy_data)`.')
|
|
return x
|
|
|
|
|
|
def get_geom_transform(self, nrows, ncols, seed=None):
|
|
"""Randomly augment a single image tensor.
|
|
|
|
# Arguments
|
|
x: 3D tensor, single image.
|
|
seed: random seed.
|
|
|
|
# Returns
|
|
A randomly transformed version of the input (same shape).
|
|
"""
|
|
|
|
if seed is not None:
|
|
np.random.seed(seed)
|
|
|
|
# use composition of homographies
|
|
# to generate final transform that needs to be applied
|
|
if self.rotation_range:
|
|
theta = np.pi / 180 * np.random.uniform(-self.rotation_range, self.rotation_range)
|
|
else:
|
|
theta = 0
|
|
|
|
if self.height_shift_range:
|
|
tx = np.random.uniform(-self.height_shift_range, self.height_shift_range) * nrows
|
|
else:
|
|
tx = 0
|
|
|
|
if self.width_shift_range:
|
|
ty = np.random.uniform(-self.width_shift_range, self.width_shift_range) * ncols
|
|
else:
|
|
ty = 0
|
|
|
|
if self.shear_range:
|
|
shear = np.random.uniform(-self.shear_range, self.shear_range)
|
|
else:
|
|
shear = 0
|
|
|
|
if self.zoom_range[0] == 1 and self.zoom_range[1] == 1:
|
|
zx, zy = 1, 1
|
|
else:
|
|
zx, zy = np.random.uniform(self.zoom_range[0], self.zoom_range[1], 2)
|
|
|
|
transform_matrix = None
|
|
if theta != 0:
|
|
rotation_matrix = np.array([[np.cos(theta), -np.sin(theta), 0],
|
|
[np.sin(theta), np.cos(theta), 0],
|
|
[0, 0, 1]])
|
|
transform_matrix = rotation_matrix
|
|
|
|
if tx != 0 or ty != 0:
|
|
shift_matrix = np.array([[1, 0, tx],
|
|
[0, 1, ty],
|
|
[0, 0, 1]])
|
|
transform_matrix = shift_matrix if transform_matrix is None else np.dot(transform_matrix, shift_matrix)
|
|
|
|
if shear != 0:
|
|
shear_matrix = np.array([[1, -np.sin(shear), 0],
|
|
[0, np.cos(shear), 0],
|
|
[0, 0, 1]])
|
|
transform_matrix = shear_matrix if transform_matrix is None else np.dot(transform_matrix, shear_matrix)
|
|
|
|
if zx != 1 or zy != 1:
|
|
zoom_matrix = np.array([[zx, 0, 0],
|
|
[0, zy, 0],
|
|
[0, 0, 1]])
|
|
transform_matrix = zoom_matrix if transform_matrix is None else np.dot(transform_matrix, zoom_matrix)
|
|
|
|
if transform_matrix is not None:
|
|
transform_matrix = transform_matrix_offset_center(transform_matrix, nrows, ncols)
|
|
return transform_matrix
|
|
|
|
def get_random_transform_specs(self, x, seed=None):
|
|
# x is a single image, so it doesn't have image number at index 0
|
|
img_row_axis = self.row_axis - 1
|
|
img_col_axis = self.col_axis - 1
|
|
img_channel_axis = self.channel_axis - 1
|
|
nrows = x.shape[img_row_axis]
|
|
ncols = x.shape[img_col_axis]
|
|
|
|
transform_matrix = self.get_geom_transform(nrows, ncols, seed=seed)
|
|
if self.horizontal_flip:
|
|
horizontal_flip = np.random.random() < 0.5
|
|
else:
|
|
horizontal_flip = False
|
|
|
|
if self.vertical_flip:
|
|
vertical_flip = np.random.random() < 0.5
|
|
else:
|
|
vertical_flip = False
|
|
return [transform_matrix, horizontal_flip, vertical_flip]
|
|
|
|
|
|
def random_transform(self, x, seed=None, fill_mode=None, cval=None,
|
|
borderMode = cv2.BORDER_TRANSPARENT,
|
|
interp = cv2.INTER_NEAREST,
|
|
use_opencv=False):
|
|
"""fill_mode and cval can be supplied to the function directly"""
|
|
if fill_mode is None:
|
|
fill_mode = self.fill_mode
|
|
if cval is None:
|
|
cval = self.cval
|
|
|
|
# x is a single image, so it doesn't have image number at index 0
|
|
img_row_axis = self.row_axis - 1
|
|
img_col_axis = self.col_axis - 1
|
|
img_channel_axis = self.channel_axis - 1
|
|
nrows = x.shape[img_row_axis]
|
|
ncols = x.shape[img_col_axis]
|
|
|
|
transform_matrix = self.get_geom_transform(nrows, ncols, seed=seed)
|
|
|
|
if transform_matrix is not None:
|
|
x = apply_affine_transform(x, transform_matrix, img_channel_axis,
|
|
fill_mode=fill_mode, cval=cval,
|
|
borderMode = borderMode,
|
|
interp = interp,
|
|
use_opencv=use_opencv)
|
|
if self.channel_shift_range != 0:
|
|
x = random_channel_shift(x,
|
|
self.channel_shift_range,
|
|
img_channel_axis)
|
|
if self.horizontal_flip:
|
|
if np.random.random() < 0.5:
|
|
x = flip_axis(x, img_col_axis)
|
|
|
|
if self.vertical_flip:
|
|
if np.random.random() < 0.5:
|
|
x = flip_axis(x, img_row_axis)
|
|
return x
|
|
|
|
def apply_transform(self, x, transform_matrix, horizontal_flip, vertical_flip,
|
|
fill_mode=None, cval=None,
|
|
borderMode = cv2.BORDER_TRANSPARENT,
|
|
interp = cv2.INTER_NEAREST,
|
|
use_opencv=False,
|
|
**kwargs):
|
|
img_channel_axis = self.channel_axis - 1
|
|
img_row_axis = self.row_axis - 1
|
|
img_col_axis = self.col_axis - 1
|
|
if transform_matrix is not None:
|
|
x = apply_affine_transform(x, transform_matrix, img_channel_axis,
|
|
fill_mode=fill_mode, cval=cval,
|
|
borderMode = borderMode,
|
|
interp = interp,
|
|
use_opencv=use_opencv)
|
|
if horizontal_flip:
|
|
x = flip_axis(x, img_col_axis)
|
|
|
|
if vertical_flip:
|
|
x = flip_axis(x, img_row_axis)
|
|
return x
|
|
|
|
|
|
def fit(self, x,
|
|
augment=False,
|
|
rounds=1,
|
|
seed=None):
|
|
"""Fits internal statistics to some sample data.
|
|
|
|
Required for featurewise_center, featurewise_std_normalization
|
|
and zca_whitening.
|
|
|
|
# Arguments
|
|
x: Numpy array, the data to fit on. Should have rank 4.
|
|
In case of grayscale data,
|
|
the channels axis should have value 1, and in case
|
|
of RGB data, it should have value 3.
|
|
augment: Whether to fit on randomly augmented samples
|
|
rounds: If `augment`,
|
|
how many augmentation passes to do over the data
|
|
seed: random seed.
|
|
|
|
# Raises
|
|
ValueError: in case of invalid input `x`.
|
|
"""
|
|
x = np.asarray(x, dtype=K.floatx())
|
|
if x.ndim != 4:
|
|
raise ValueError('Input to `.fit()` should have rank 4. '
|
|
'Got array with shape: ' + str(x.shape))
|
|
if x.shape[self.channel_axis] not in {1, 3, 4}:
|
|
warnings.warn(
|
|
'Expected input to be images (as Numpy array) '
|
|
'following the data format convention "' + self.data_format + '" '
|
|
'(channels on axis ' + str(self.channel_axis) + '), i.e. expected '
|
|
'either 1, 3 or 4 channels on axis ' + str(self.channel_axis) + '. '
|
|
'However, it was passed an array with shape ' + str(x.shape) +
|
|
' (' + str(x.shape[self.channel_axis]) + ' channels).')
|
|
|
|
if seed is not None:
|
|
np.random.seed(seed)
|
|
|
|
x = np.copy(x)
|
|
if augment:
|
|
ax = np.zeros(tuple([rounds * x.shape[0]] + list(x.shape)[1:]), dtype=K.floatx())
|
|
for r in range(rounds):
|
|
for i in range(x.shape[0]):
|
|
ax[i + r * x.shape[0]] = self.random_transform(x[i])
|
|
x = ax
|
|
|
|
if self.featurewise_center:
|
|
self.mean = np.mean(x, axis=(0, self.row_axis, self.col_axis))
|
|
broadcast_shape = [1, 1, 1]
|
|
broadcast_shape[self.channel_axis - 1] = x.shape[self.channel_axis]
|
|
self.mean = np.reshape(self.mean, broadcast_shape)
|
|
x -= self.mean
|
|
|
|
if self.featurewise_std_normalization:
|
|
self.std = np.std(x, axis=(0, self.row_axis, self.col_axis))
|
|
broadcast_shape = [1, 1, 1]
|
|
broadcast_shape[self.channel_axis - 1] = x.shape[self.channel_axis]
|
|
self.std = np.reshape(self.std, broadcast_shape)
|
|
x /= (self.std + K.epsilon())
|
|
|
|
if self.zca_whitening:
|
|
flat_x = np.reshape(x, (x.shape[0], x.shape[1] * x.shape[2] * x.shape[3]))
|
|
sigma = np.dot(flat_x.T, flat_x) / flat_x.shape[0]
|
|
u, s, _ = linalg.svd(sigma)
|
|
self.principal_components = np.dot(np.dot(u, np.diag(1. / np.sqrt(s + self.zca_epsilon))), u.T)
|
|
|
|
|
|
class Iterator(Sequence):
|
|
"""Abstract base class for image data iterators.
|
|
|
|
# Arguments
|
|
n: Integer, total number of samples in the dataset to loop over.
|
|
batch_size: Integer, size of a batch.
|
|
shuffle: Boolean, whether to shuffle the data between epochs.
|
|
seed: Random seeding for data shuffling.
|
|
stratify: Label to stratify by
|
|
oversampling: Oversample if stratification label is provided to match the multiples of the largest class
|
|
"""
|
|
|
|
def __init__(self, n, batch_size, shuffle, seed,
|
|
postprocessing_function=None, stratify=None, oversampling=True,
|
|
subsample_factor=None,
|
|
subsample_num=None,
|
|
batch_rate=1,
|
|
):
|
|
self.stratify = stratify
|
|
self.batch_rate = batch_rate
|
|
self.oversampling = oversampling
|
|
self.n = n
|
|
self.batch_size = batch_size
|
|
self.seed = seed
|
|
self.shuffle = shuffle
|
|
self.batch_index = 0
|
|
self.total_batches_seen = 0
|
|
self.lock = threading.Lock()
|
|
self.index_array = None
|
|
self.postprocessing_function = postprocessing_function
|
|
self.index_generator = self._flow_index()
|
|
self.final_num = self.n
|
|
if self.stratify is not None:
|
|
self._prep_stratified()
|
|
if self.oversampling:
|
|
self.final_num = sum(np.asarray(list(self.class_size.values()))>0) * max(self.class_size.values())
|
|
elif (subsample_factor is not None) or (subsample_num is not None):
|
|
if subsample_num is None:
|
|
self.final_num = int(self.n/subsample_factor)
|
|
else:
|
|
print("subsample_num", subsample_num)
|
|
self.final_num = subsample_num
|
|
self.orig_index_array = np.random.randint(0, self.n, size=self.final_num)
|
|
else:
|
|
self.orig_index_array = np.arange(self.n)
|
|
|
|
def _prep_stratified(self):
|
|
self.uniq_classes = np.unique(self.stratify)
|
|
#self.class_size = np.bincount(self.stratify)
|
|
self.class_size = Counter(self.stratify)
|
|
|
|
self.class_inds = {}
|
|
for cc in self.uniq_classes:
|
|
mask = np.asarray(self.stratify) == cc
|
|
self.class_inds[cc] = np.where(mask)[0]
|
|
|
|
#ex_per_class = max(self.class_size)
|
|
ex_per_class = max(self.class_size.values())
|
|
self.orig_index_array = []
|
|
#for cc,ss in enumerate(self.class_size):
|
|
for cc in self.uniq_classes:
|
|
ss = self.class_size[cc]
|
|
if ss==0:
|
|
continue
|
|
self.orig_index_array.append(
|
|
np.random.choice(self.class_inds[cc],
|
|
size=ex_per_class,
|
|
replace=ss<ex_per_class)
|
|
)
|
|
# np.random.permutation(self.n)
|
|
self.orig_index_array = np.stack(self.orig_index_array).T.ravel()
|
|
|
|
def _set_index_array(self):
|
|
self.index_array = self.orig_index_array.copy()
|
|
if self.shuffle:
|
|
if self.stratify is not None and not self.oversampling:
|
|
self.index_array = np.random.choice(self.orig_index_array.copy(),
|
|
size=self.final_num, replace=False)
|
|
else:
|
|
self.index_array = np.random.permutation(self.orig_index_array.copy())
|
|
|
|
def __getitem__(self, idx):
|
|
if idx >= len(self):
|
|
raise ValueError('Asked to retrieve element {idx}, '
|
|
'but the Sequence '
|
|
'has length {length}'.format(idx=idx,
|
|
length=len(self)))
|
|
if self.seed is not None:
|
|
np.random.seed(self.seed + self.total_batches_seen)
|
|
self.total_batches_seen += 1
|
|
if self.index_array is None:
|
|
self._set_index_array()
|
|
_batch_size = self.batch_size // self.batch_rate
|
|
index_array = self.index_array[_batch_size * idx:
|
|
_batch_size * (idx + 1)]
|
|
return self._get_batches_of_transformed_samples(index_array)
|
|
|
|
def __len__(self):
|
|
if hasattr(self, "batch_rate"):
|
|
_batch_size = self.batch_size // self.batch_rate
|
|
else:
|
|
_batch_size = self.batch_size
|
|
return int(np.ceil(self.n / float(_batch_size)))
|
|
|
|
def on_epoch_end(self):
|
|
self._set_index_array()
|
|
|
|
def reset(self):
|
|
self.batch_index = 0
|
|
|
|
def _flow_index(self):
|
|
# Ensure self.batch_index is 0.
|
|
self.reset()
|
|
if hasattr(self, "batch_rate"):
|
|
_batch_size = self.batch_size // self.batch_rate
|
|
else:
|
|
_batch_size = self.batch_size
|
|
|
|
while 1:
|
|
if self.seed is not None:
|
|
np.random.seed(self.seed + self.total_batches_seen)
|
|
if self.batch_index == 0:
|
|
self._set_index_array()
|
|
|
|
self.batch_index = self.total_batches_seen % self.__len__()
|
|
current_index = (self.batch_index * _batch_size) % self.final_num
|
|
self.total_batches_seen += 1
|
|
yield self.index_array[current_index:
|
|
current_index + _batch_size]
|
|
|
|
def __iter__(self):
|
|
# Needed if we want to do something like:
|
|
# for x, y in data_gen.flow(...):
|
|
return self
|
|
|
|
def __next__(self, *args, **kwargs):
|
|
return self.next(*args, **kwargs)
|
|
|
|
|
|
class NumpyArrayIterator(Iterator):
|
|
"""Iterator yielding data from a Numpy array.
|
|
|
|
# Arguments
|
|
x: Numpy array of input data.
|
|
y: Numpy array of targets data.
|
|
image_data_generator: Instance of `ImageDataGenerator`
|
|
to use for random transformations and normalization.
|
|
batch_size: Integer, size of a batch.
|
|
shuffle: Boolean, whether to shuffle the data between epochs.
|
|
seed: Random seed for data shuffling.
|
|
data_format: String, one of `channels_first`, `channels_last`.
|
|
save_to_dir: Optional directory where to save the pictures
|
|
being yielded, in a viewable format. This is useful
|
|
for visualizing the random transformations being
|
|
applied, for debugging purposes.
|
|
save_prefix: String prefix to use for saving sample
|
|
images (if `save_to_dir` is set).
|
|
save_format: Format to use for saving sample images
|
|
(if `save_to_dir` is set).
|
|
"""
|
|
|
|
def __init__(self, x, y, image_data_generator,
|
|
batch_size=32, shuffle=False, seed=None,
|
|
data_format=None,
|
|
save_to_dir=None, save_prefix='', save_format='png',
|
|
postprocessing_function=None,
|
|
color_mode=None,
|
|
stratify=None,
|
|
oversampling=True
|
|
):
|
|
channels_axis = 3 if data_format == 'channels_last' else 1
|
|
self.channels_axis = channels_axis
|
|
if y is not None and len(x) != len(y):
|
|
raise ValueError('X (images tensor) and y (labels) '
|
|
'should have the same length. '
|
|
'Found: X.shape = %s, y.shape = %s' %
|
|
(np.asarray(x).shape, np.asarray(y).shape))
|
|
|
|
if data_format is None:
|
|
data_format = K.image_data_format()
|
|
self.x = np.asarray(x, dtype=K.floatx())
|
|
|
|
if self.x.ndim != 4:
|
|
warnings.warn('Input data in `NumpyArrayIterator` '
|
|
'should have rank 4. You passed an array '
|
|
'with shape\t%s' % str(self.x.shape))
|
|
else:
|
|
if self.x.shape[channels_axis] not in {1, 3, 4}:
|
|
warnings.warn('NumpyArrayIterator is set to use the '
|
|
'data format convention "' + data_format + '" '
|
|
'(channels on axis ' + str(channels_axis) + '), i.e. expected '
|
|
'either 1, 3 or 4 channels on axis ' + str(channels_axis) + '. '
|
|
'However, it was passed an array with shape ' + str(self.x.shape) +
|
|
' (' + str(self.x.shape[channels_axis]) + ' channels).')
|
|
if y is not None:
|
|
self.y = np.asarray(y)
|
|
else:
|
|
self.y = None
|
|
self.image_data_generator = image_data_generator
|
|
self.data_format = data_format
|
|
self.save_to_dir = save_to_dir
|
|
self.save_prefix = save_prefix
|
|
self.save_format = save_format
|
|
self.color_mode=color_mode
|
|
super(NumpyArrayIterator, self).__init__(x.shape[0], batch_size, shuffle, seed,
|
|
stratify=stratify, oversampling=oversampling,
|
|
postprocessing_function=postprocessing_function)
|
|
print("self.color_mode", self.color_mode)
|
|
|
|
|
|
def _get_batches_of_transformed_samples(self, index_array):
|
|
print("self.batch_index", self.batch_index,)
|
|
print("==", index_array)
|
|
batch_x = np.zeros(tuple([len(index_array)] + list(self.x.shape)[1:]),
|
|
dtype=K.floatx())
|
|
if len(batch_x.shape)==3:
|
|
batch_x = batch_x.reshape(batch_x.shape + (1,))
|
|
for i, j in enumerate(index_array):
|
|
x = self.x[j]
|
|
if len(x.shape)==2:
|
|
x = x.reshape(x.shape + (1,))
|
|
x = self.image_data_generator.random_transform(x.astype(K.floatx()))
|
|
x = self.image_data_generator.standardize(x)
|
|
batch_x[i] = x
|
|
if self.save_to_dir:
|
|
for i, j in enumerate(index_array):
|
|
img = array_to_img(batch_x[i], self.data_format, scale=True)
|
|
fname = '{prefix}_{index}_{hash}.{format}'.format(prefix=self.save_prefix,
|
|
index=j,
|
|
hash=np.random.randint(1e4),
|
|
format=self.save_format)
|
|
img.save(os.path.join(self.save_to_dir, fname))
|
|
if self.color_mode in (3,'rgb'):
|
|
if len(batch_x.shape)==3:
|
|
batch_x = np.stack([ batch_x ]*3, axis=-1)
|
|
if batch_x.shape[self.channels_axis]==1:
|
|
batch_x = np.concatenate([ batch_x ]*3, axis=3)
|
|
#print("batch_x", batch_x.shape)
|
|
#raise Exception("test!!!")
|
|
if self.postprocessing_function is not None:
|
|
batch_x = self.postprocessing_function(batch_x)
|
|
|
|
if self.y is None:
|
|
return batch_x
|
|
batch_y = self.y[index_array]
|
|
return batch_x, batch_y
|
|
|
|
def next(self):
|
|
"""For python 2.x.
|
|
|
|
# Returns
|
|
The next batch.
|
|
"""
|
|
# Keeps under lock only the mechanism which advances
|
|
# the indexing of each batch.
|
|
with self.lock:
|
|
index_array = next(self.index_generator)
|
|
# The transformation of images is not under thread lock
|
|
# so it can be done in parallel
|
|
return self._get_batches_of_transformed_samples(index_array)
|
|
|
|
|
|
def _count_valid_files_in_directory(directory, white_list_formats, follow_links):
|
|
"""Count files with extension in `white_list_formats` contained in a directory.
|
|
|
|
# Arguments
|
|
directory: absolute path to the directory containing files to be counted
|
|
white_list_formats: set of strings containing allowed extensions for
|
|
the files to be counted.
|
|
|
|
# Returns
|
|
the count of files with extension in `white_list_formats` contained in
|
|
the directory.
|
|
"""
|
|
def _recursive_list(subpath):
|
|
return sorted(os.walk(subpath, followlinks=follow_links), key=lambda tpl: tpl[0])
|
|
|
|
samples = 0
|
|
for root, _, files in _recursive_list(directory):
|
|
for fname in files:
|
|
is_valid = False
|
|
for extension in white_list_formats:
|
|
if fname.lower().endswith('.' + extension):
|
|
is_valid = True
|
|
break
|
|
if is_valid:
|
|
samples += 1
|
|
return samples
|
|
|
|
|
|
def _list_valid_filenames_in_directory(directory, white_list_formats,
|
|
class_indices, follow_links):
|
|
"""List paths of files in `subdir` relative from `directory` whose extensions are in `white_list_formats`.
|
|
|
|
# Arguments
|
|
directory: absolute path to a directory containing the files to list.
|
|
The directory name is used as class label and must be a key of `class_indices`.
|
|
white_list_formats: set of strings containing allowed extensions for
|
|
the files to be counted.
|
|
class_indices: dictionary mapping a class name to its index.
|
|
|
|
# Returns
|
|
classes: a list of class indices
|
|
filenames: the path of valid files in `directory`, relative from
|
|
`directory`'s parent (e.g., if `directory` is "dataset/class1",
|
|
the filenames will be ["class1/file1.jpg", "class1/file2.jpg", ...]).
|
|
"""
|
|
def _recursive_list(subpath):
|
|
return sorted(os.walk(subpath, followlinks=follow_links), key=lambda tpl: tpl[0])
|
|
|
|
classes = []
|
|
filenames = []
|
|
subdir = os.path.basename(directory)
|
|
basedir = os.path.dirname(directory)
|
|
for root, _, files in _recursive_list(directory):
|
|
for fname in sorted(files):
|
|
is_valid = False
|
|
for extension in white_list_formats:
|
|
if fname.lower().endswith('.' + extension):
|
|
is_valid = True
|
|
break
|
|
if is_valid:
|
|
classes.append(class_indices[subdir])
|
|
# add filename relative to directory
|
|
absolute_path = os.path.join(root, fname)
|
|
filenames.append(os.path.relpath(absolute_path, basedir))
|
|
return classes, filenames
|
|
|
|
|
|
class DirectoryIterator(Iterator):
|
|
"""Iterator capable of reading images from a directory on disk.
|
|
|
|
# Arguments
|
|
directory: Path to the directory to read images from.
|
|
Each subdirectory in this directory will be
|
|
considered to contain images from one class,
|
|
or alternatively you could specify class subdirectories
|
|
via the `classes` argument.
|
|
image_data_generator: Instance of `ImageDataGenerator`
|
|
to use for random transformations and normalization.
|
|
target_size: tuple of integers, dimensions to resize input images to.
|
|
color_mode: One of `"rgb"`, `"grayscale"`. Color mode to read images.
|
|
classes: Optional list of strings, names of subdirectories
|
|
containing images from each class (e.g. `["dogs", "cats"]`).
|
|
It will be computed automatically if not set.
|
|
class_mode: Mode for yielding the targets:
|
|
`"binary"`: binary targets (if there are only two classes),
|
|
`"categorical"`: categorical targets,
|
|
`"sparse"`: integer targets,
|
|
`"input"`: targets are images identical to input images (mainly
|
|
used to work with autoencoders),
|
|
`None`: no targets get yielded (only input images are yielded).
|
|
batch_size: Integer, size of a batch.
|
|
shuffle: Boolean, whether to shuffle the data between epochs.
|
|
seed: Random seed for data shuffling.
|
|
data_format: String, one of `channels_first`, `channels_last`.
|
|
save_to_dir: Optional directory where to save the pictures
|
|
being yielded, in a viewable format. This is useful
|
|
for visualizing the random transformations being
|
|
applied, for debugging purposes.
|
|
save_prefix: String prefix to use for saving sample
|
|
images (if `save_to_dir` is set).
|
|
save_format: Format to use for saving sample images
|
|
(if `save_to_dir` is set).
|
|
"""
|
|
|
|
def __init__(self, directory, image_data_generator,
|
|
target_size=(256, 256), color_mode='rgb',
|
|
classes=None, class_mode='categorical',
|
|
batch_size=32, shuffle=True, seed=None,
|
|
data_format=None,
|
|
save_to_dir=None, save_prefix='', save_format='png',
|
|
postprocessing_function=None,
|
|
follow_links=False,
|
|
stratify=None, oversampling=True,
|
|
subsample_factor= None,
|
|
subsample_num = None,
|
|
output_filenames=None,
|
|
):
|
|
|
|
self.output_filenames=output_filenames
|
|
#self.postprocessing_function = postprocessing_function
|
|
if data_format is None:
|
|
data_format = K.image_data_format()
|
|
self.directory = directory
|
|
self.image_data_generator = image_data_generator
|
|
self.target_size = tuple(target_size)
|
|
if color_mode not in {'rgb', 'grayscale'}:
|
|
raise ValueError('Invalid color mode:', color_mode,
|
|
'; expected "rgb" or "grayscale".')
|
|
self.color_mode = color_mode
|
|
self.data_format = data_format
|
|
if self.color_mode == 'rgb':
|
|
if self.data_format == 'channels_last':
|
|
self.image_shape = self.target_size + (3,)
|
|
else:
|
|
self.image_shape = (3,) + self.target_size
|
|
else:
|
|
if self.data_format == 'channels_last':
|
|
self.image_shape = self.target_size + (1,)
|
|
else:
|
|
self.image_shape = (1,) + self.target_size
|
|
self.classes = classes
|
|
if class_mode not in {'categorical', 'binary', 'sparse',
|
|
'input', None}:
|
|
raise ValueError('Invalid class_mode:', class_mode,
|
|
'; expected one of "categorical", '
|
|
'"binary", "sparse", "input"'
|
|
' or None.')
|
|
self.class_mode = class_mode
|
|
self.save_to_dir = save_to_dir
|
|
self.save_prefix = save_prefix
|
|
self.save_format = save_format
|
|
|
|
white_list_formats = {'png', 'jpg', 'jpeg', 'bmp', 'ppm'}
|
|
|
|
# first, count the number of samples and classes
|
|
self.samples = 0
|
|
|
|
if not classes:
|
|
classes = []
|
|
for subdir in sorted(os.listdir(directory)):
|
|
if os.path.isdir(os.path.join(directory, subdir)):
|
|
classes.append(subdir)
|
|
self.num_class = len(classes)
|
|
self.class_indices = dict(zip(classes, range(len(classes))))
|
|
|
|
pool = multiprocessing.pool.ThreadPool()
|
|
function_partial = partial(_count_valid_files_in_directory,
|
|
white_list_formats=white_list_formats,
|
|
follow_links=follow_links)
|
|
self.samples = sum(pool.map(function_partial,
|
|
(os.path.join(directory, subdir)
|
|
for subdir in classes)))
|
|
|
|
print('Found %d images belonging to %d classes.' % (self.samples, self.num_class))
|
|
|
|
# second, build an index of the images in the different class subfolders
|
|
results = []
|
|
|
|
self.filenames = []
|
|
self.classes = np.zeros((self.samples,), dtype='int32')
|
|
i = 0
|
|
for dirpath in (os.path.join(directory, subdir) for subdir in classes):
|
|
results.append(pool.apply_async(_list_valid_filenames_in_directory,
|
|
(dirpath, white_list_formats,
|
|
self.class_indices, follow_links)))
|
|
for res in results:
|
|
classes, filenames = res.get()
|
|
self.classes[i:i + len(classes)] = classes
|
|
self.filenames += filenames
|
|
i += len(classes)
|
|
pool.close()
|
|
pool.join()
|
|
super(DirectoryIterator, self).__init__(self.samples, batch_size, shuffle, seed,
|
|
stratify=self.classes if stratify else None,
|
|
oversampling=oversampling,
|
|
subsample_factor=subsample_factor,
|
|
subsample_num=subsample_num,
|
|
postprocessing_function=postprocessing_function)
|
|
|
|
def _get_batches_of_transformed_samples(self, index_array):
|
|
batch_x = np.zeros((len(index_array),) + self.image_shape, dtype=K.floatx())
|
|
batch_fn = []
|
|
grayscale = self.color_mode == 'grayscale'
|
|
# build batch of image data
|
|
for i, j in enumerate(index_array):
|
|
fname = self.filenames[j]
|
|
batch_fn.append(fname)
|
|
|
|
img = load_img(os.path.join(self.directory, fname),
|
|
grayscale=grayscale,
|
|
target_size=self.target_size)
|
|
x = img_to_array(img, data_format=self.data_format)
|
|
|
|
x = self.image_data_generator.random_transform(x)
|
|
x = self.image_data_generator.standardize(x)
|
|
batch_x[i] = x
|
|
if self.postprocessing_function:
|
|
batch_x = self.postprocessing_function(batch_x)
|
|
# optionally save augmented images to disk for debugging purposes
|
|
if self.save_to_dir:
|
|
for i, j in enumerate(index_array):
|
|
img = array_to_img(batch_x[i], self.data_format, scale=True)
|
|
fname = '{prefix}_{index}_{hash}.{format}'.format(prefix=self.save_prefix,
|
|
index=j,
|
|
hash=np.random.randint(1e4),
|
|
format=self.save_format)
|
|
img.save(os.path.join(self.save_to_dir, fname))
|
|
# build batch of labels
|
|
if self.class_mode == 'input':
|
|
batch_y = batch_x.copy()
|
|
elif self.class_mode == 'sparse':
|
|
batch_y = self.classes[index_array]
|
|
elif self.class_mode == 'binary':
|
|
batch_y = self.classes[index_array].astype(K.floatx())
|
|
elif self.class_mode == 'categorical':
|
|
batch_y = np.zeros((len(batch_x), self.num_class), dtype=K.floatx())
|
|
for i, label in enumerate(self.classes[index_array]):
|
|
batch_y[i, label] = 1.
|
|
else:
|
|
return batch_x
|
|
if self.output_filenames:
|
|
return batch_x, batch_y, batch_fn
|
|
return batch_x, batch_y
|
|
|
|
def next(self):
|
|
"""For python 2.x.
|
|
|
|
# Returns
|
|
The next batch.
|
|
"""
|
|
with self.lock:
|
|
index_array = next(self.index_generator)
|
|
# The transformation of images is not under thread lock
|
|
# so it can be done in parallel
|
|
return self._get_batches_of_transformed_samples(index_array)
|
|
|
|
# coding: utf-8
|
|
#from keras.utils.data_utils import Sequence
|
|
#import numpy as np
|
|
from numpy.lib.format import open_memmap
|
|
def get_slice(center, size=(256,256),
|
|
target_size = (None, None),
|
|
reflect = False,
|
|
):
|
|
size = np.asarray(size)
|
|
xy = np.asarray(center) - size//2
|
|
end = xy + size
|
|
|
|
margins_max = []
|
|
margins_min = []
|
|
slice_ = []
|
|
for s_, e_, t_, sz in zip(xy, end, target_size, size):
|
|
if t_:
|
|
m_e = max(0, -(t_ - e_ ))
|
|
e_ = min(e_, t_)
|
|
else:
|
|
m_e = None
|
|
m_s = max(0, -s_)
|
|
s_ = max(s_,0)
|
|
if reflect and (m_s>0) or (m_e>0):
|
|
#print("before", s_,e_, m_s, m_e)
|
|
if (e_==0) and m_e > sz//2:
|
|
m_e, s_ = t_-s_, t_ - m_e
|
|
if (s_==0) and (m_e> e_):
|
|
e_, m_s = m_s, e_
|
|
#print("after", s_,e_, m_s, m_e)
|
|
margins_max.append(m_e)
|
|
margins_min.append(m_s)
|
|
slice_.append(slice(s_, e_))
|
|
return slice_, list(zip(margins_min, margins_max))
|
|
|
|
|
|
def pad_patch(img, slc, padseq):
|
|
"""if no padding required, returns a slice (pointer);
|
|
otherwise returns a padded copy of a slice
|
|
|
|
INPUT:
|
|
- target_size [ width, height ]
|
|
"""
|
|
if any((s > 0 for s in padseq[0])) | any((s > 0 for s in padseq[1])):
|
|
patch = np.pad(img[slc], padseq, mode='constant')
|
|
else:
|
|
patch = img[slc]
|
|
return patch
|
|
|
|
class PatchIterator(Iterator):
|
|
"""
|
|
point_sampler(filename, label=2)
|
|
"""
|
|
def __init__(self, fn_img, fn_pnt,
|
|
point_sampler,
|
|
image_data_generator = None,
|
|
batch_size = 4,
|
|
shuffle=True,
|
|
seed=0,
|
|
patch_size = (512,512),
|
|
dtype='uint16',
|
|
mode = 'reflect',
|
|
label_freq={1:5, 2:10},
|
|
augmentation=None,
|
|
postprocessing_functions = [None, None],
|
|
color_mode = None,
|
|
output_indices = False,
|
|
patches_per_image = 1,
|
|
):
|
|
self.fn_img = fn_img
|
|
self.fn_pnt = fn_pnt
|
|
assert batch_size % patches_per_image == 0, (
|
|
"batch_size must be multiple of patches_per_image")
|
|
self.patches_per_image = patches_per_image
|
|
self.imgs_per_batch = batch_size // self.patches_per_image
|
|
assert len(fn_img) == len(fn_pnt)
|
|
assert len(fn_img) >0
|
|
print("%d images supplied" % len(fn_img))
|
|
self.point_sampler = point_sampler
|
|
nsamples = len(self.fn_img)
|
|
self.mode = mode
|
|
self.postprocessing_functions = postprocessing_functions
|
|
self.color_mode=color_mode
|
|
self._reflect = self.mode == 'reflect'
|
|
self.augmentation = augmentation
|
|
self.image_data_generator = image_data_generator
|
|
self.output_indices = output_indices
|
|
self.transforms = []
|
|
if self.image_data_generator is not None:
|
|
self.transforms.append( self.image_data_generator.random_transform )
|
|
self.transforms.append( self.image_data_generator.standardize )
|
|
|
|
_norm_const = sum(label_freq.values())
|
|
#print(_norm_const)
|
|
self.patch_size = tuple(patch_size)
|
|
self.dtype = dtype
|
|
self.label_freq_dict = {kk:vv/_norm_const for kk,vv in label_freq.items()}
|
|
self.labels = np.asarray(list(self.label_freq_dict.keys()))
|
|
self.label_freq = list(self.label_freq_dict.values())
|
|
self.label_cum_freq = np.cumsum(self.label_freq)
|
|
|
|
self.index_generator = self._flow_index()
|
|
super(PatchIterator, self).__init__(nsamples, batch_size, shuffle, seed,
|
|
batch_rate = patches_per_image,
|
|
# stratify=stratify, oversampling=oversampling,
|
|
# postprocessing_function=postprocessing_function
|
|
)
|
|
def sample_label(self, batch_size=None):
|
|
if batch_size == None:
|
|
batch_size = self.batch_size
|
|
matr = np.random.rand(batch_size, self.label_cum_freq.shape[0])>= self.label_cum_freq
|
|
label_inds = np.argmin(matr, axis=1)
|
|
return np.asarray([self.labels[x] for x in label_inds])
|
|
|
|
def init_indices():
|
|
self.indices = np.random.randint(len(self.fn_img))
|
|
|
|
def open_npy(self):
|
|
for ff in self.fn_img:
|
|
img = open_memmap(ff, dtype=np.uint16, mode='r', shape=target_size[::-1])
|
|
|
|
def sample_points(self, index, labels):
|
|
fm = self.fn_pnt[index]
|
|
try:
|
|
samplegen = self.point_sampler(fm, labels)
|
|
for pt, label in zip(samplegen, labels):
|
|
yield (label, pt)
|
|
#pt = mzl.point(fm, buf=None, label=2, position=-1)
|
|
except OSError as ee:
|
|
print("error on ", fm)
|
|
raise ee
|
|
|
|
def sample_img(self, img, pt, buffer=None, extend_dim=False, transforms=[]):
|
|
shape = img.shape
|
|
slc, padseq = get_slice(pt, size=self.patch_size, target_size=shape, reflect= self._reflect)
|
|
if buffer is None:
|
|
patch = pad_patch(img, slc, padseq)
|
|
return patch
|
|
else:
|
|
if self.mode == 'constant':
|
|
outslc = [slice(ss.start-ss.start, ss.stop-ss.start) for ss in slc]
|
|
if extend_dim:
|
|
outslc += [0]
|
|
buffer[outslc] = img[slc]
|
|
else:
|
|
outslc = [slice(None)]*len(slc)
|
|
if extend_dim:
|
|
outslc += [0]
|
|
buffer[outslc] = np.pad(img[slc], padseq, self.mode)#[:,:]
|
|
if self.augmentation and len(transforms)>0:
|
|
for tt in transforms:
|
|
if buffer is None:
|
|
print(fi, pt, tt)
|
|
buffer[:] = tt(buffer.copy())[:]
|
|
#print("finally after rescale", buffer.max())
|
|
|
|
def _get_batches_of_transformed_samples(self, sample_inds):
|
|
if sample_inds is None:
|
|
with self.lock:
|
|
sample_inds = next(self.index_generator)
|
|
# Repeat each image index
|
|
# sample_inds = list(itertools.chain.from_iterable(itertools.repeat(x, self.imgs_per_batch) for x in sample_inds))
|
|
#batch_class_inds = self.sample_label()
|
|
curr_batch_size = len(sample_inds) * self.patches_per_image
|
|
pts = np.zeros((curr_batch_size, 2), dtype='uint16')
|
|
slices = [None, slice(None), slice(None)]
|
|
if self.color_mode is None:
|
|
extend_dim = False
|
|
buffer = np.zeros((curr_batch_size, ) + self.patch_size, dtype= self.dtype)
|
|
elif self.color_mode in ('grayscale', 'greyscale', 1):
|
|
extend_dim = True
|
|
buffer = np.zeros((curr_batch_size, ) + self.patch_size + (1,), dtype= self.dtype)
|
|
else:
|
|
raise ValueError("`color_mode` should be None or 'grayscale'")
|
|
|
|
batch_class_inds = np.zeros(curr_batch_size, dtype='uint16')
|
|
for nn, (ss) in enumerate(sample_inds):
|
|
labels = self.sample_label(self.patches_per_image)
|
|
lbl_slice = slice(self.patches_per_image*nn, self.patches_per_image*(nn+1))
|
|
batch_class_inds[lbl_slice] = labels
|
|
points = self.sample_points(ss, labels)
|
|
img = open_memmap(self.fn_img[ss], mode='r',)
|
|
for jj, (lbl, pt) in enumerate(points):
|
|
#range(self.patches_per_image):
|
|
#print("nn", nn, "jj", jj, "pt", pt)
|
|
pts[jj, :] = pt
|
|
slices[0] = nn * self.patches_per_image + jj
|
|
#print("buffer size", buffer.shape)
|
|
#print("slice", slices[0])
|
|
self.sample_img(img, pt, buffer[slices],
|
|
transforms=self.transforms,
|
|
extend_dim=extend_dim)
|
|
|
|
output = [buffer, batch_class_inds]
|
|
if self.output_indices:
|
|
output.append(sample_inds)
|
|
output.append(pts)
|
|
|
|
for ii, (dd, ff) in enumerate(zip(output, self.postprocessing_functions)):
|
|
# print(ii, ff)
|
|
if ff is not None:
|
|
output[ii] = ff(dd)
|
|
return output #, pts
|
|
|
|
# pool = multiprocessing.pool.ThreadPool()
|
|
|
|
def next(self):
|
|
"""For python 2.x.
|
|
|
|
# Returns
|
|
The next batch.
|
|
"""
|
|
with self.lock:
|
|
index_array = next(self.index_generator)
|
|
# The transformation of images is not under thread lock
|
|
# so it can be done in parallel
|
|
return self._get_batches_of_transformed_samples(index_array)#import threading
|
|
|
|
class MemMapDataset():
|
|
def __init__(self, root_dir, csv_file,
|
|
classes = ["Control", "Case",],
|
|
label_col = "label",
|
|
filename_col = "filename",
|
|
binary=True,
|
|
transform=None,
|
|
nsamples = None,
|
|
encode_label = None):
|
|
import pandas as pd
|
|
self.transform = transform
|
|
self.csv_file = csv_file
|
|
self.table = pd.read_csv(csv_file)
|
|
if nsamples:
|
|
self.table = self.table[:nsamples]
|
|
self.filenames = self.table[filename_col].tolist()
|
|
self.classes = self.table[label_col].tolist()
|
|
self.root_dir = root_dir
|
|
self.label_col = label_col
|
|
self.filename_col = filename_col
|
|
for cc in classes:
|
|
if cc not in self.classes:
|
|
print("class is missing in data: %s" % cc)
|
|
#classes = np.unique(self.classes).tolist()
|
|
self.class_set = classes
|
|
if encode_label is None:
|
|
if len(classes) in (1,2) and binary:
|
|
self.onehot = MemMapDataset.encode_label_binary(self.table[label_col], self.class_set)
|
|
else:
|
|
self.onehot = MemMapDataset.encode_label_onehot(self.table[label_col], self.class_set)
|
|
else:
|
|
self.onehot = encode_label(self.table[label_col], self.class_set)
|
|
|
|
@staticmethod
|
|
def encode_label_binary(labelvector, class_set):
|
|
onehot = np.stack([(labelvector == cc).values for cc in class_set[1:]], axis=-1)
|
|
return onehot
|
|
|
|
@staticmethod
|
|
def encode_label_onehot(labelvector, class_set):
|
|
onehot = np.stack([(labelvector == cc).values for cc in class_set], axis=-1)
|
|
return onehot
|
|
|
|
def __len__(self):
|
|
return len(self.table)
|
|
|
|
def __getitem__(self, idx):
|
|
item = self.table.iloc[idx]
|
|
img_name = os.path.join(self.root_dir, item[self.filename_col])
|
|
|
|
image = open_memmap(img_name, mode='r')
|
|
label = self.onehot[idx]
|
|
|
|
if self.transform:
|
|
sample = self.transform([image, label])
|
|
else:
|
|
sample = [image, label]
|
|
|
|
return sample
|
|
|
|
class MemMapIterator(Iterator):
|
|
def __init__(self, root_dir, csv_file,
|
|
classes = ["Control", "Case",],
|
|
image_data_generator = None,
|
|
binary=True,
|
|
transform=None,
|
|
nsamples = None,
|
|
batch_size = 1,
|
|
shuffle = False,
|
|
seed = None,
|
|
postprocessing_function=None,
|
|
stratify=None,
|
|
oversampling=True,
|
|
subsample_factor=None,
|
|
subsample_num=None,
|
|
batch_rate=1,
|
|
dtype = K.floatx(),
|
|
color_mode=None,
|
|
data_format = 'channels_last',
|
|
label_col = "label",
|
|
filename_col = "filename",
|
|
encode_label=None,
|
|
):
|
|
channels_axis = 3 if data_format == 'channels_last' else 1
|
|
self.channels_axis = channels_axis
|
|
self.dtype = dtype
|
|
self.color_mode=color_mode
|
|
#self.image_data_generator = image_data_generator
|
|
self.transforms = []
|
|
if image_data_generator is not None:
|
|
self.transforms.append( image_data_generator.random_transform )
|
|
self.transforms.append( image_data_generator.standardize )
|
|
|
|
self.dataset = MemMapDataset(root_dir, csv_file, classes=classes, binary=binary,
|
|
transform=transform, nsamples=nsamples,
|
|
label_col = label_col,
|
|
filename_col = filename_col,
|
|
encode_label=encode_label,
|
|
)
|
|
self.classes = self.dataset.classes
|
|
self.filenames = self.dataset.filenames
|
|
|
|
super(MemMapIterator, self).__init__(len(self.dataset), batch_size, shuffle, seed,
|
|
stratify=self.classes if stratify else None,
|
|
oversampling=oversampling,
|
|
subsample_factor=subsample_factor,
|
|
subsample_num=subsample_num,
|
|
postprocessing_function=postprocessing_function,
|
|
)
|
|
|
|
def _get_batches_of_transformed_samples(self, index_array):
|
|
print("index_array", index_array)
|
|
batch_x = np.zeros(tuple([len(index_array)] + list(self.dataset[0][0].shape)),
|
|
dtype=self.dtype)
|
|
if len(batch_x.shape)==3:
|
|
batch_x = batch_x.reshape(batch_x.shape + (1,))
|
|
for i, j in enumerate(index_array):
|
|
x = self.dataset[j][0]
|
|
if len(x.shape)==2:
|
|
x = x.reshape(x.shape + (1,))
|
|
x = x.astype(self.dtype)
|
|
for tt in self.transforms:
|
|
x = tt(x)
|
|
batch_x[i] = x
|
|
if self.color_mode in (3,'rgb'):
|
|
if len(batch_x.shape)==3:
|
|
batch_x = np.stack([ batch_x ]*3, axis=-1)
|
|
if batch_x.shape[self.channels_axis]==1:
|
|
batch_x = np.concatenate([ batch_x ]*3, axis=3)
|
|
#print("batch_x", batch_x.shape)
|
|
#raise Exception("test!!!")
|
|
if self.postprocessing_function is not None:
|
|
batch_x = self.postprocessing_function(batch_x)
|
|
if len(self.dataset[0])==1:
|
|
return batch_x
|
|
batch_y = np.asarray([self.dataset[j][1] for j in index_array])
|
|
return batch_x, batch_y
|
|
|
|
def next(self):
|
|
"""For python 2.x.
|
|
|
|
# Returns
|
|
The next batch.
|
|
"""
|
|
# Keeps under lock only the mechanism which advances
|
|
# the indexing of each batch.
|
|
with self.lock:
|
|
index_array = next(self.index_generator)
|
|
# The transformation of images is not under thread lock
|
|
# so it can be done in parallel
|
|
return self._get_batches_of_transformed_samples(index_array)
|
|
|
|
|
|
def read_decode_coco(fname):
|
|
with open(fname) as fh:
|
|
coco = json.load(fh)
|
|
return decode(coco)
|
|
|
|
class MemMapCocoDataset():
|
|
def __init__(self, root_dir, csv_file,
|
|
binary=True,
|
|
transform=None,
|
|
nsamples = None,
|
|
mmapcol = 'memmap',
|
|
cococol = 'coco',
|
|
):
|
|
import pandas as pd
|
|
self.transform = transform
|
|
self.table = pd.read_csv(csv_file)
|
|
if nsamples:
|
|
self.table = self.table[:nsamples]
|
|
self.root_dir = root_dir
|
|
self.mmapcol = mmapcol
|
|
self.cococol = cococol
|
|
|
|
def __len__(self):
|
|
return len(self.table)
|
|
|
|
def __getitem__(self, idx):
|
|
item = self.table.iloc[idx]
|
|
|
|
img_name = os.path.join(self.root_dir,
|
|
item[self.mmapcol])
|
|
image = open_memmap(img_name, mode='r')
|
|
|
|
coco_name = os.path.join(self.root_dir,
|
|
item[self.cococol])
|
|
label = read_decode_coco(coco_name)
|
|
|
|
if self.transform:
|
|
sample = self.transform([image, label])
|
|
else:
|
|
sample = [image, label]
|
|
|
|
return sample
|
|
|
|
|
|
def resize_inputs(xx, yy,
|
|
mode='constant',
|
|
target_size = [512, 512],
|
|
constant_values_x = 255,
|
|
constant_values_y = 0,
|
|
add_const_to_label = 0,
|
|
):
|
|
xx = (crop_pad_center(xx, target_size, pad_mode=mode, constant_values=constant_values_x))
|
|
if add_const_to_label>0:
|
|
yy += add_const_to_label
|
|
yy = (crop_pad_center(yy, target_size, pad_mode=mode, constant_values=constant_values_y))
|
|
if len(yy.shape) == 2:
|
|
yy= yy[..., np.newaxis]
|
|
return xx, yy
|
|
|
|
class MemMapCocoIterator(Iterator):
|
|
def __init__(self, root_dir, csv_file,
|
|
image_data_generator = None,
|
|
binary=True,
|
|
mode='constant',
|
|
target_size = [512, 512],
|
|
constant_values_x = 255,
|
|
constant_values_y = 0,
|
|
nsamples = None,
|
|
batch_size = 1,
|
|
shuffle = False,
|
|
seed = None,
|
|
postprocessing_function=None,
|
|
stratify=None,
|
|
oversampling=True,
|
|
subsample_factor=None,
|
|
subsample_num=None,
|
|
batch_rate=1,
|
|
dtype = K.floatx(),
|
|
color_mode=None,
|
|
data_format = 'channels_last',
|
|
output_indices = False,
|
|
add_const_to_label=0,
|
|
):
|
|
self.output_indices = output_indices
|
|
channels_axis = 3 if data_format == 'channels_last' else 1
|
|
self.channels_axis = channels_axis
|
|
self.dtype = dtype
|
|
self.color_mode=color_mode
|
|
#self.image_data_generator = image_data_generator
|
|
#self.transforms = []
|
|
if image_data_generator is not None:
|
|
self.get_random_transform_specs = image_data_generator.get_random_transform_specs
|
|
self.apply_geom_transform = image_data_generator.apply_transform
|
|
self.intensity_transform = image_data_generator.standardize
|
|
#self.transforms.append( image_data_generator.random_transform )
|
|
#self.transforms.append( image_data_generator.standardize )
|
|
if target_size is not None:
|
|
transform = lambda x : resize_inputs(x[0], x[1],
|
|
mode=mode,
|
|
target_size = target_size,
|
|
constant_values_x = constant_values_x,
|
|
constant_values_y = constant_values_y,
|
|
add_const_to_label=add_const_to_label,
|
|
)
|
|
self.constant_values_x = constant_values_x
|
|
self.constant_values_y = constant_values_y
|
|
|
|
self.dataset = MemMapCocoDataset(root_dir, csv_file, binary=binary,
|
|
transform=transform, nsamples=nsamples)
|
|
|
|
super(MemMapCocoIterator, self).__init__(len(self.dataset), batch_size, shuffle, seed,
|
|
stratify=stratify, oversampling=oversampling,
|
|
subsample_factor=subsample_factor,
|
|
subsample_num=subsample_num,
|
|
postprocessing_function=postprocessing_function
|
|
)
|
|
|
|
def _get_batches_of_transformed_samples(self, index_array):
|
|
#import ipdb
|
|
#ipdb.set_trace()
|
|
batch_x = np.zeros(tuple([len(index_array)] + list(self.dataset[0][0].shape)),
|
|
dtype=self.dtype)
|
|
batch_y = np.zeros(tuple([len(index_array)] + list(self.dataset[0][1].shape)),
|
|
dtype=self.dtype)
|
|
if len(batch_x.shape)==3:
|
|
batch_x = batch_x.reshape(batch_x.shape + (1,))
|
|
for i, j in enumerate(index_array):
|
|
x, y = self.dataset[j]
|
|
if len(x.shape)==2:
|
|
x = x.reshape(x.shape + (1,))
|
|
|
|
if hasattr(self, 'get_random_transform_specs'):
|
|
transform_matrix, horizontal_flip, vertical_flip = self.get_random_transform_specs(x)
|
|
x = self.apply_geom_transform(x, transform_matrix, horizontal_flip, vertical_flip,
|
|
interp=cv2.INTER_CUBIC,
|
|
#interp=cv2.INTER_NEAREST,
|
|
borderMode = cv2.BORDER_CONSTANT,
|
|
cval = [self.constant_values_x] *3,
|
|
use_opencv=True)
|
|
y = self.apply_geom_transform(y, transform_matrix, horizontal_flip, vertical_flip,
|
|
interp=cv2.INTER_NEAREST,
|
|
borderMode = cv2.BORDER_CONSTANT,
|
|
cval=self.constant_values_y,
|
|
use_opencv=True)
|
|
if hasattr(self, 'intensity_transform'):
|
|
x = self.intensity_transform(x)
|
|
x = x.astype(self.dtype)
|
|
batch_x[i] = x
|
|
batch_y[i] = y
|
|
if self.color_mode in (3,'rgb'):
|
|
if len(batch_x.shape)==3:
|
|
batch_x = np.stack([ batch_x ]*3, axis=-1)
|
|
if batch_x.shape[self.channels_axis]==1:
|
|
batch_x = np.concatenate([ batch_x ]*3, axis=3)
|
|
#print("batch_x", batch_x.shape)
|
|
#raise Exception("test!!!")
|
|
if self.postprocessing_function is not None:
|
|
batch_x = self.postprocessing_function(batch_x)
|
|
if self.output_indices:
|
|
if len(self.dataset[0])==1:
|
|
return batch_x, index_array
|
|
return batch_x, batch_y, index_array
|
|
else:
|
|
if len(self.dataset[0])==1:
|
|
return batch_x
|
|
return batch_x, batch_y
|
|
|
|
def next(self):
|
|
"""For python 2.x.
|
|
|
|
# Returns
|
|
The next batch.
|
|
"""
|
|
# Keeps under lock only the mechanism which advances
|
|
# the indexing of each batch.
|
|
with self.lock:
|
|
index_array = next(self.index_generator)
|
|
# The transformation of images is not under thread lock
|
|
# so it can be done in parallel
|
|
return self._get_batches_of_transformed_samples(index_array)
|