File size: 34,503 Bytes
747451d | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 |
# /*---------------------------------------------------------------------------------------------
# * Copyright (c) 2022-2023 STMicroelectronics.
# * All rights reserved.
# *
# * This software is licensed under terms that can be found in the LICENSE file in
# * the root directory of this software component.
# * If no LICENSE file comes with this software, it is provided AS-IS.
# *--------------------------------------------------------------------------------------------*/
"""
References:
----------
Some of the code in this package is from or was inspired by:
Keras Image Preprocessing Layers
The Tensorflow Authors
Copyright (c) 2019
Link to the source code:
https://github.com/keras-team/keras/blob/v2.12.0/keras/layers/preprocessing/image_preprocessing.py#L394-L495
"""
import math
import tensorflow as tf
from common.data_augmentation import \
check_fill_and_interpolation, transform_images, check_dataaug_argument, \
get_flip_matrix, get_translation_matrix, get_rotation_matrix, \
get_shear_matrix, get_zoom_matrix
from .objdet_random_utils import objdet_apply_change_rate
def _transform_boxes(boxes, transforms, image_width, image_height, scale=1.):
"""
This function applies affine transformations to a batch of boxes.
The transformation matrices are independent from each other
and are generally different from one batch item to another.
Arguments:
boxes:
Boxes the matrices are applied to
Shape:[batch_size, num_boxes, 4]
transforms:
Matrices coefficients to apply to the boxes
Shape:[batch_size, 8]
Returns:
Transformed boxes
Shape:[batch_size, num_boxes, 4]
"""
image_width = tf.cast(image_width, tf.float32)
image_height = tf.cast(image_height, tf.float32)
boxes_shape = tf.shape(boxes)
batch_size = boxes_shape[0]
num_boxes = boxes_shape[1]
# Create a mask to keep track of padding boxes
coords_sum = tf.math.reduce_sum(boxes, axis=-1)
padding_mask = tf.where(coords_sum > 0, 1., 0.)
padding_mask = tf.repeat(padding_mask, 4)
padding_mask = tf.reshape(padding_mask, [batch_size, num_boxes, 4])
# Create and invert the matrices (inversion is necessary
# to align with the TF function that transforms the images)
transforms = tf.concat([
transforms,
tf.ones([batch_size, 1], dtype=tf.float32)],
axis=-1)
matrices = tf.reshape(transforms, [batch_size, 3, 3])
matrices = tf.linalg.inv(matrices)
# The same transform has to be applied to all the boxes
# of a batch item, so we replicate the matrices.
matrices = tf.expand_dims(matrices, axis=1)
matrices = tf.tile(matrices, [1, num_boxes, 1, 1])
x1 = boxes[..., 0]
y1 = boxes[..., 1]
x2 = boxes[..., 2]
y2 = boxes[..., 3]
# Reduce the size of the boxes before transforming them
if scale < 1:
dx = scale * (x2 - x1)
dy = scale * (y2 - y1)
boxes = tf.stack([x1 + dx, y1 + dx, x2 - dx, y2 - dy], axis=-1)
# Stack box corner vectors to create 4x4 matrices
# Then multiply by transformation matrices to get
# the transformed corner vectors.
corners = tf.concat([
tf.stack([x1, x2, x2, x1], axis=-1),
tf.stack([y1, y1, y2, y2], axis=-1),
tf.ones([batch_size, num_boxes, 4], dtype=tf.float32)],
axis=-1)
corners = tf.reshape(corners, [batch_size, num_boxes, 3, 4])
trd_corners = tf.linalg.matmul(matrices, corners)
# Project transformed corner vectors onto x and y axis
tx1 = tf.math.reduce_min(trd_corners[..., 0, :], axis=-1)
tx2 = tf.math.reduce_max(trd_corners[..., 0, :], axis=-1)
ty1 = tf.math.reduce_min(trd_corners[..., 1, :], axis=-1)
ty2 = tf.math.reduce_max(trd_corners[..., 1, :], axis=-1)
# Clip transformed coordinates
tx1 = tf.math.maximum(tx1, 0)
tx1 = tf.math.minimum(tx1, image_width)
tx2 = tf.math.maximum(tx2, 0)
tx2 = tf.math.minimum(tx2, image_width)
ty1 = tf.math.maximum(ty1, 0)
ty1 = tf.math.minimum(ty1, image_height)
ty2 = tf.math.maximum(ty2, 0)
ty2 = tf.math.minimum(ty2, image_height)
trd_boxes = tf.stack([tx1, ty1, tx2, ty2], axis=-1)
# Get rid of boxes that don't make sense
valid_boxes = tf.math.logical_and(tx2 > tx1, ty2 > ty1)
valid_boxes = tf.cast(valid_boxes, tf.float32)
trd_boxes *= tf.expand_dims(valid_boxes, axis=-1)
# Set to 0 the coordinates of padding boxes as transforms
# may have resulted in some non-zeros coordinates.
trd_boxes *= padding_mask
return trd_boxes
#------------------------- Random flip -------------------------
def objdet_random_flip(images, labels, mode=None, change_rate=0.5):
"""
This function randomly flips input images and the bounding boxes
in the associated groundtruth labels.
Setting `change_rate` to 0.5 usually gives good results (don't set
it to 1.0, otherwise all the images will be flipped).
Arguments:
images:
Input RGB or grayscale images
Shape: [batch_size, width, height, channels]
labels:
Groundtruth labels associated to the images in
(class, x1, y1, x2, y2) format. Bounding box coordinates
must be absolute, opposite corners coordinates.
Shape: [batch_size, num_labels, 5]
mode:
A string representing the flip axis. Either "horizontal",
"vertical" or "horizontal_and_vertical".
change_rate:
A float in the interval [0, 1] representing the number of
changed images versus the total number of input images average
ratio. For example, if `change_rate` is set to 0.25, 25% of
the input images will get changed on average (75% won't get
changed). If it is set to 0.0, no images are changed. If it is
set to 1.0, all the images are changed.
Returns:
The flipped images and groundtruth labels with flipped bounding boxes.
"""
if mode not in ("horizontal", "vertical", "horizontal_and_vertical"):
raise ValueError(
"Argument `mode` of function `random_flip`: supported values are 'horizontal', "
"'vertical' and 'horizontal_and_vertical'. Received {}".format(mode))
images_shape = tf.shape(images)
batch_size = images_shape[0]
image_width = images_shape[1]
image_height = images_shape[2]
matrix = get_flip_matrix(batch_size, image_width, image_height, mode)
boxes = labels[..., 1:]
flipped_images = transform_images(images, matrix)
flipped_boxes = _transform_boxes(boxes, matrix, image_width, image_height)
# Apply the change rate to images and labels
images_aug, boxes_aug = objdet_apply_change_rate(
images, boxes, flipped_images, flipped_boxes, change_rate=change_rate)
classes = tf.expand_dims(labels[..., 0], axis=-1)
labels_aug = tf.concat([classes, boxes_aug], axis=-1)
return images_aug, labels_aug
#------------------------- Random translation -------------------------
def objdet_random_translation(
images, labels,
width_factor, height_factor,
fill_mode='reflect', interpolation='bilinear', fill_value=0.0,
change_rate=1.0):
"""
This function randomly translates input images and the bounding boxes
in the associated groundtruth labels.
Arguments:
images:
Input RGB or grayscale images with shape
Shape: [batch_size, width, height, channels]
labels:
Groundtruth labels associated to the images in
(class, x1, y1, x2, y2) format. Bounding box coordinates
must be absolute, opposite corners coordinates.
Shape: [batch_size, num_labels, 5]
width_factor:
A float or a tuple of 2 floats, specifies the range of values
the horizontal shift factors are sampled from (one per image).
If a scalar value v is used, it is equivalent to the tuple (-v, v).
A negative factor means shifting the image left, while a positive
factor means shifting the image right.
For example, `width_factor`=(-0.2, 0.3) results in an output shifted
left by up to 20% or shifted right by up to 30%.
height_factor:
A float or a tuple of 2 floats, specifies the range of values
the vertical shift factors are sampled from (one per image).
If a scalar value v is used, it is equivalent to the tuple (-v, v).
A negative factor means shifting the image up, while a positive
factor means shifting the image down.
For example, `height_factor`=(-0.2, 0.3) results in an output shifted
up by up to 20% or shifted down by up to 30%.
fill_mode:
Points outside the boundaries of the input are filled according
to the given mode. One of {'constant', 'reflect', 'wrap', 'nearest'}.
See Tensorflow documentation at https://tensorflow.org
for more details.
interpolation:
A string, the interpolation method. Supported values: 'nearest', 'bilinear'.
change_rate:
A float in the interval [0, 1] representing the number of
changed images versus the total number of input images average
ratio. For example, if `change_rate` is set to 0.25, 25% of
the input images will get changed on average (75% won't get
changed). If it is set to 0.0, no images are changed. If it is
set to 1.0, all the images are changed.
Returns:
The translated images and groundtruth labels with translated bounding boxes.
"""
check_dataaug_argument(width_factor, "width_factor", function_name="random_translation", data_type=float)
if isinstance(width_factor, (tuple, list)):
width_lower = width_factor[0]
width_upper = width_factor[1]
else:
width_lower = -width_factor
width_upper = width_factor
check_dataaug_argument(height_factor, "height_factor", function_name="random_translation", data_type=float)
if isinstance(height_factor, (tuple, list)):
height_lower = height_factor[0]
height_upper = height_factor[1]
else:
height_lower = -height_factor
height_upper = height_factor
check_fill_and_interpolation(fill_mode, interpolation, fill_value, function_name="random_translation")
images_shape = tf.shape(images)
batch_size = images_shape[0]
image_width = images_shape[1]
image_height = images_shape[2]
classes = labels[..., 0]
boxes = labels[..., 1:]
width_translate = tf.random.uniform(
[batch_size, 1], minval=width_lower, maxval=width_upper, dtype=tf.float32)
width_translate = width_translate * tf.cast(image_width, tf.float32)
height_translate = tf.random.uniform(
[batch_size, 1], minval=height_lower, maxval=height_upper, dtype=tf.float32)
height_translate = height_translate * tf.cast(image_height, tf.float32)
translations = tf.cast(
tf.concat([width_translate, height_translate], axis=1),
dtype=tf.float32)
translation_matrix = get_translation_matrix(translations)
translated_images = transform_images(
images,
translation_matrix,
interpolation=interpolation,
fill_mode=fill_mode,
fill_value=fill_value)
translated_boxes = _transform_boxes(
boxes,
translation_matrix,
image_width,
image_height)
# Apply the change rate to images and labels
images_aug, boxes_aug = objdet_apply_change_rate(
images, boxes, translated_images, translated_boxes, change_rate=change_rate)
classes = tf.expand_dims(labels[..., 0], axis=-1)
labels_aug = tf.concat([classes, boxes_aug], axis=-1)
return images_aug, labels_aug
#------------------------- Random rotation -------------------------
def objdet_random_rotation(
images, labels, factor=None,
fill_mode='reflect', interpolation='bilinear', fill_value=0.0,
change_rate=1.0):
"""
This function randomly rotates input images and the bounding boxes
in the associated groundtruth labels.
Arguments:
images:
Input RGB or grayscale images with shape
Shape: [batch_size, width, height, channels]
labels:
Groundtruth labels associated to the images in
(class, x1, y1, x2, y2) format. Bounding box coordinates
must be absolute, opposite corners coordinates.
Shape: [batch_size, num_labels, 5]
factor:
A float or a tuple of 2 floats, specifies the range of values the
rotation angles are sampled from (one per image). If a scalar
value v is used, it is equivalent to the tuple (-v, v).
Rotation angles are in gradients (fractions of 2*pi). A positive
angle means rotating counter clock-wise, while a negative angle
means rotating clock-wise.
For example, `factor`=(-0.2, 0.3) results in an output rotated by
a random amount in the range [-20% * 2pi, 30% * 2pi].
fill_mode:
Points outside the boundaries of the input are filled according
to the given mode. One of {'constant', 'reflect', 'wrap', 'nearest'}.
See Tensorflow documentation at https://tensorflow.org
for more details.
interpolation:
A string, the interpolation method. Supported values: 'nearest', 'bilinear'.
change_rate:
A float in the interval [0, 1] representing the number of
changed images versus the total number of input images average
ratio. For example, if `change_rate` is set to 0.25, 25% of
the input images will get changed on average (75% won't get
changed). If it is set to 0.0, no images are changed. If it is
set to 1.0, all the images are changed.
Returns:
The rotated images and groundtruth labels with rotated bounding boxes.
"""
check_dataaug_argument(factor, "factor", function_name="random_rotation", data_type=float)
if not isinstance(factor, (tuple, list)):
factor = (-factor, factor)
check_fill_and_interpolation(fill_mode, interpolation, fill_value, function_name="random_rotation")
images_shape = tf.shape(images)
batch_size = images_shape[0]
image_width = images_shape[1]
image_height = images_shape[2]
min_angle = factor[0] * 2. * math.pi
max_angle = factor[1] * 2. * math.pi
angles = tf.random.uniform([batch_size], minval=min_angle, maxval=max_angle)
classes = labels[..., 0]
boxes = labels[..., 1:]
rotation_matrix = get_rotation_matrix(angles, image_width, image_height)
rotated_images = transform_images(
images,
rotation_matrix,
fill_mode=fill_mode,
fill_value=fill_value,
interpolation=interpolation)
rotated_boxes = _transform_boxes(
boxes,
rotation_matrix,
image_width,
image_height,
scale=0.1)
# Apply the change rate to images and labels
images_aug, boxes_aug = objdet_apply_change_rate(
images, boxes, rotated_images, rotated_boxes, change_rate=change_rate)
classes = tf.expand_dims(labels[..., 0], axis=-1)
labels_aug = tf.concat([classes, boxes_aug], axis=-1)
return images_aug, labels_aug
#------------------------- Random shear -------------------------
def objdet_random_shear(
images,
labels,
factor=None,
axis='xy',
fill_mode='reflect',
interpolation='bilinear',
fill_value=0.0,
change_rate=1.0):
"""
This function randomly shears input images.
Arguments:
images:
Input RGB or grayscale images with shape
[batch_size, width, height, channels].
factor:
A float or a tuple of 2 floats, specifies the range of values
the shear angles are sampled from (one per image). If a scalar
value v is used, it is equivalent to the tuple (-v, v). Angles
are in radians (fractions of 2*pi).
For example, factor=(-0.349, 0.785) results in an output sheared
by a random angle in the range [-20 degrees, +45 degrees].
axis:
The shear axis:
'xy': shear along both axis
'x': shear along the x axis only
'y': shear along the y axis only
fill_mode:
Points outside the boundaries of the input are filled according
to the given mode. One of {'constant', 'reflect', 'wrap', 'nearest'}.
See Tensorflow documentation at https://tensorflow.org
for more details.
interpolation:
A string, the interpolation method. Supported values: 'nearest', 'bilinear'.
change_rate:
A float in the interval [0, 1] representing the number of
changed images versus the total number of input images average
ratio. For example, if `change_rate` is set to 0.25, 25% of
the input images will get changed on average (75% won't get
changed). If it is set to 0.0, no images are changed. If it is
set to 1.0, all the images are changed.
Returns:
The sheared images.
"""
if axis == 'x':
function_name = "random_shear_x"
elif axis == 'y':
function_name = "random_shear_y"
else:
function_name = "random_shear"
check_dataaug_argument(factor, "factor", function_name=function_name, data_type=float)
if not isinstance(factor, (tuple, list)):
factor = (-factor, factor)
check_fill_and_interpolation(fill_mode, interpolation, fill_value, function_name=function_name)
images_shape = tf.shape(images)
batch_size = images_shape[0]
image_width = images_shape[1]
image_height = images_shape[2]
min_angle = factor[0] * 2. * math.pi
max_angle = factor[1] * 2. * math.pi
angles = tf.random.uniform([batch_size], minval=min_angle, maxval=max_angle)
classes = labels[..., 0]
boxes = labels[..., 1:]
shear_matrix = get_shear_matrix(angles, axis=axis)
sheared_images = transform_images(
images,
shear_matrix,
fill_mode=fill_mode,
fill_value=fill_value,
interpolation=interpolation)
sheared_boxes = _transform_boxes(
boxes,
shear_matrix,
image_width,
image_height,
scale=0.1)
# Apply the change rate to images and labels
images_aug, boxes_aug = objdet_apply_change_rate(
images, boxes, sheared_images, sheared_boxes, change_rate=change_rate)
classes = tf.expand_dims(labels[..., 0], axis=-1)
labels_aug = tf.concat([classes, boxes_aug], axis=-1)
return images_aug, labels_aug
#------------------------- Random zoom -------------------------
def objdet_random_zoom(
images, labels, width_factor=None, height_factor=None,
fill_mode='reflect', interpolation='bilinear', fill_value=0.0,
change_rate=1.0):
"""
This function randomly zooms input images and the bounding boxes
in the associated groundtruth labels.
If `width_factor` and `height_factor` are both set, the images are zoomed
in or out on each axis independently, which may result in noticeable distortion.
If you want to avoid distortion, only set `width_factor` and the mages will be
zoomed by the same amount in both directions.
Arguments:
images:
Input RGB or grayscale images with shape
Shape: [batch_size, width, height, channels]
labels:
Groundtruth labels associated to the images in
(class, x1, y1, x2, y2) format. Bounding box coordinates
must be absolute, opposite corners coordinates.
Shape: [batch_size, num_labels, 5]
width_factor:
A float or a tuple of 2 floats, specifies the range of values horizontal
zoom factors are sampled from (one per image). If a scalar value v is used,
it is equivalent to the tuple (-v, v). Factors are fractions of the width
of the image. A positive factor means zooming out, while a negative factor
means zooming in.
For example, width_factor=(0.2, 0.3) results in an output zoomed out by
a random amount in the range [+20%, +30%]. width_factor=(-0.3, -0.2) results
in an output zoomed in by a random amount in the range [+20%, +30%].
height_factor:
A float or a tuple of 2 floats, specifies the range of values vertical
zoom factors are sampled from (one per image). If a scalar value v is used,
it is equivalent to the tuple (-v, v). Factors are fractions of the height
of the image. A positive value means zooming out, while a negative value
means zooming in.
For example, height_factor=(0.2, 0.3) results in an output zoomed out
between 20% to 30%. height_factor=(-0.3, -0.2) results in an output zoomed
in between 20% to 30%.
If `height_factor` is not set, it defaults to None. In this case, images
images will be zoomed by the same amounts in both directions and no image
distortion will occur.
fill_mode:
Points outside the boundaries of the input are filled according
to the given mode. One of {'constant', 'reflect', 'wrap', 'nearest'}.
See Tensorflow documentation at https://tensorflow.org
for more details.
interpolation:
A string, the interpolation method. Supported values: 'nearest', 'bilinear'.
change_rate:
A float in the interval [0, 1] representing the number of
changed images versus the total number of input images average
ratio. For example, if `change_rate` is set to 0.25, 25% of
the input images will get changed on average (75% won't get
changed). If it is set to 0.0, no images are changed. If it is
set to 1.0, all the images are changed.
Returns:
The zoomed images and groundtruth labels with zoomed bounding boxes.
"""
check_dataaug_argument(width_factor, "width_factor", function_name="random_zoom", data_type=float)
if isinstance(width_factor, (tuple, list)):
width_lower = width_factor[0]
width_upper = width_factor[1]
else:
width_lower = -width_factor
width_upper = width_factor
if height_factor is not None:
check_dataaug_argument(height_factor, "height_factor", function_name="random_zoom", data_type=float)
if isinstance(height_factor, (tuple, list)):
height_lower = height_factor[0]
height_upper = height_factor[1]
else:
height_lower = -height_factor
height_upper = height_factor
if abs(height_lower) > 1.0 or abs(height_upper) > 1.0:
raise ValueError(
"Argument `height_factor` of function `random_zoom`: expecting float "
"values in the interval [-1.0, 1.0]. Received: {}".format(height_factor))
else:
height_lower = width_lower
height_upper = width_upper
check_fill_and_interpolation(fill_mode, interpolation, fill_value, function_name="random_zoom")
images_shape = tf.shape(images)
batch_size = images_shape[0]
image_width = images_shape[1]
image_height = images_shape[2]
classes = labels[..., 0]
boxes = labels[..., 1:]
height_zoom = tf.random.uniform(
[batch_size, 1], minval=1. + height_lower, maxval=1. + height_upper, dtype=tf.float32)
width_zoom = tf.random.uniform(
[batch_size, 1], minval=1. + width_lower, maxval=1. + width_upper, dtype=tf.float32)
zooms = tf.cast(tf.concat([width_zoom, height_zoom], axis=1), dtype=tf.float32)
zoom_matrix = get_zoom_matrix(zooms, image_width, image_height)
zoomed_images = transform_images(
images,
zoom_matrix,
fill_mode=fill_mode,
fill_value=fill_value,
interpolation=interpolation)
zoomed_boxes = _transform_boxes(
boxes,
zoom_matrix,
image_width,
image_height)
# Apply the change rate to images and labels
images_aug, boxes_aug = objdet_apply_change_rate(
images, boxes, zoomed_images, zoomed_boxes, change_rate=change_rate)
classes = tf.expand_dims(labels[..., 0], axis=-1)
labels_aug = tf.concat([classes, boxes_aug], axis=-1)
return images_aug, labels_aug
#--------------------------------- Random bounded crop ---------------------
def objdet_random_bounded_crop(
images,
labels,
width_factor=None,
height_factor=None,
crop_center_x=None,
crop_center_y=None,
fill_mode='reflect',
interpolation='bilinear',
fill_value=0.0,
change_rate=1.0):
"""
This function randomly crops or dezoom on each axis of the input images.
If `width_factor` and `height_factor` are both set, the images are zoomed
in or out on each axis independently, which may result in noticeable distortion.
If you want to avoid distortion, only set `width_factor` and the mages will be
zoomed by the same amount in both directions.
Arguments:
images:
Input RGB or grayscale images with shape
[batch_size, width, height, channels].
labels:
Groundtruth labels associated to the images in
(class, x1, y1, x2, y2) format. Bounding box coordinates
must be absolute, opposite corners coordinates.
Shape: [batch_size, num_labels, 5]
width_factor:
A float or a tuple of 2 floats, specifies the range of values horizontal
zoom factors are sampled from (one per image). If a scalar value v is used,
it is equivalent to the tuple (-v, v). Factors are fractions of the width
of the image. A positive factor means zooming out, while a negative factor
means zooming in.
For example, width_factor=(0.2, 0.3) results in an output zoomed out by
a random amount in the range [+20%, +30%]. width_factor=(-0.3, -0.2) results
in an output zoomed in by a random amount in the range [+20%, +30%].
height_factor:
A float or a tuple of 2 floats, specifies the range of values vertical
zoom factors are sampled from (one per image). If a scalar value v is used,
it is equivalent to the tuple (-v, v). Factors are fractions of the height
of the image. A positive value means zooming out, while a negative value
means zooming in.
For example, height_factor=(0.2, 0.3) results in an output zoomed out
between 20% to 30%. height_factor=(-0.3, -0.2) results in an output zoomed
in between 20% to 30%.
If `height_factor` is not set, it defaults to None. In this case, images
images will be zoomed by the same amounts in both directions and no image
distortion will occur.
fill_mode:
Points outside the boundaries of the input are filled according
to the given mode. One of {'constant', 'reflect', 'wrap', 'nearest'}.
See Tensorflow documentation at https://tensorflow.org
for more details.
interpolation:
A string, the interpolation method. Supported values: 'nearest', 'bilinear'.
change_rate:
A float in the interval [0, 1] representing the number of
changed images versus the total number of input images average
ratio. For example, if `change_rate` is set to 0.25, 25% of
the input images will get changed on average (75% won't get
changed). If it is set to 0.0, no images are changed. If it is
set to 1.0, all the images are changed.
Returns:
The zoomed images and groundtruth labels with zoomed bounding boxes.
"""
check_dataaug_argument(width_factor, "width_factor", function_name="random_bounded_crop", data_type=float)
if isinstance(width_factor, (tuple, list)):
width_lower = width_factor[0]
width_upper = width_factor[1]
else:
width_lower = width_factor
width_upper = width_factor
if height_factor is not None:
check_dataaug_argument(height_factor, "height_factor", function_name="random_bounded_crop", data_type=float)
if isinstance(height_factor, (tuple, list)):
height_lower = height_factor[0]
height_upper = height_factor[1]
else:
height_lower = height_factor
height_upper = height_factor
if abs(height_lower) > 1.0 or abs(height_upper) > 1.0:
raise ValueError(
"Argument `height_factor` of function `random_bounded_crop`: expecting float "
"values in the interval [-1.0, 1.0]. Received: {}".format(height_factor))
if (crop_center_x is not None) and (crop_center_y is not None):
check_dataaug_argument(crop_center_x, "crop_center_x", function_name="random_bounded_crop", data_type=float)
if isinstance(crop_center_x, (tuple, list)) and isinstance(crop_center_y, (tuple, list)):
t_width_lower = 0.5 - crop_center_x[1]
t_width_upper = 0.5 - crop_center_x[0]
t_height_lower = 0.5 - crop_center_y[1]
t_height_upper = 0.5 - crop_center_y[0]
else:
t_width_lower = - crop_center_x/2
t_width_upper = crop_center_x/2
t_height_lower = -crop_center_y/2
t_height_upper = crop_center_y/2
if abs(t_width_lower) > 1.0 or abs(t_width_upper) > 1.0 or abs(t_height_lower) > 1.0 or abs(t_height_upper) > 1.0:
raise ValueError(
"Argument `crop_center_x` or `crop_center_y` of function `random_bounded_crop`: expecting float "
"values in the interval [-1.0, 1.0]. Received: {}, {}".format(crop_center_x,crop_center_y))
else:
t_width_lower = None
t_width_upper = None
t_height_lower = None
t_height_upper = None
check_fill_and_interpolation(fill_mode, interpolation, fill_value, function_name="random_bounded_crop")
image_shape = tf.shape(images)
batch_size = image_shape[0]
width = tf.cast(image_shape[1], tf.float32)
height = tf.cast(image_shape[2], tf.float32)
classes = labels[..., 0]
boxes = labels[..., 1:]
zoom_width = tf.random.uniform(
[batch_size, 1], minval=width_lower, maxval=width_upper, dtype=tf.float32)
if height_factor is not None:
zoom_height = tf.random.uniform(
[batch_size, 1], minval=height_lower, maxval=height_upper, dtype=tf.float32)
else:
zoom_height = zoom_width
zoom_factor_w = 1-zoom_width
zoom_factor_h = 1-zoom_height
zoom_factor_w *= tf.cast(zoom_factor_w>=0,tf.float32)
zoom_factor_h *= tf.cast(zoom_factor_h>=0,tf.float32)
if t_width_lower is None:
translation_width = tf.random.uniform(
[batch_size, 1], minval=-1, maxval=1, dtype=tf.float32)
translation_height = tf.random.uniform(
[batch_size, 1], minval=-1, maxval=1, dtype=tf.float32)
translation_width *= zoom_factor_w/2
translation_height *= zoom_factor_h/2
else:
translation_width = tf.random.uniform(
[batch_size, 1], minval=t_width_lower, maxval=t_width_upper, dtype=tf.float32)
translation_height = tf.random.uniform(
[batch_size, 1], minval=t_height_lower, maxval=t_height_upper, dtype=tf.float32)
zooms = tf.cast(tf.concat([zoom_width, zoom_height], axis=1), dtype=tf.float32) # shape : (batch, 2)
translations = tf.cast(
tf.concat([translation_width * width, translation_height * height], axis=1),
dtype=tf.float32)
zoom_matrix = get_zoom_matrix(zooms, width, height)
translation_matrix = get_translation_matrix(translations)
translated_images = transform_images(
images,
translation_matrix,
fill_mode=fill_mode,
fill_value=fill_value,
interpolation=interpolation)
translated_boxes = _transform_boxes(
boxes,
translation_matrix,
width,
height)
zoomed_images = transform_images(
translated_images,
zoom_matrix,
fill_mode=fill_mode,
fill_value=fill_value,
interpolation=interpolation)
zoomed_boxes = _transform_boxes(
translated_boxes,
zoom_matrix,
width,
height)
# Apply the change rate to images and labels
images_aug, boxes_aug = objdet_apply_change_rate(
images, boxes, zoomed_images, zoomed_boxes, change_rate=change_rate)
classes = tf.expand_dims(labels[..., 0], axis=-1)
labels_aug = tf.concat([classes, boxes_aug], axis=-1)
return images_aug, labels_aug |