def preprocess_for_test(image, out_shape=IMAGE_SIZE, scope='ssd_preprocessing_test'):
with tf.name_scope(scope):
image = tf.to_float(image)
image = tf_image.tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
image = tf_image.resize_image(image, out_shape)
bbox_img = tf.constant([0., 0., 1., 1.])
return image, bbox_img
def preprocess_for_train(image,
labels,
bboxes,
height,
width,
out_shape,
data_format='NHWC',
use_whiten=True,
scope='textbox_process_train'):
"""Preprocesses the given image for training.
Args:
image: A `Tensor` representing an image of arbitrary size.
labels : A Tensor inlcudes all labels
bboxes : A Tensor inlcudes cordinates of bbox in shape [N, 4]
out_shape : Image_size ,default is [300, 300]
Returns:
A preprocessed image.
"""
with tf.name_scope(scope, 'textbox_process_train',
[image, labels, bboxes]):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
bboxes = tf.minimum(bboxes, 1.0)
bboxes = tf.maximum(bboxes, 0.0)
image = tf_image.distorter(image)
image, bboxes = tf_image.resize_image_bboxes_with_crop_or_pad2(
image, bboxes, height[0], width[0])
image, labels, bboxes = tf_image.Random_crop(image, labels, bboxes)
image = tf_image.resize_image(image,
out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
image = tf.clip_by_value(image, 0., 1.)
image, bboxes = tf_image.random_flip_left_right(image, bboxes)
num = tf.reduce_sum(tf.cast(labels, tf.int32))
image.set_shape([out_shape[0], out_shape[1], 3])
tf_image.tf_summary_image(image, bboxes)
image = image * 255.
image = tf_image.tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
bboxes = tf.minimum(bboxes, 1.0)
bboxes = tf.maximum(bboxes, 0.0)
#image = tf.subtract(image, 128.)
#image = tf.multiply(image, 2.0)
if data_format == 'NHWC':
image = image
else:
image = tf.transpose(image, perm=(2, 0, 1))
return image, labels, bboxes, num
def preprocess_for_train(image, labels, bboxes,
out_shape, data_format='NHWC', use_whiten=True,
scope='textbox_process_train'):
with tf.name_scope(scope, 'textbox_process_train', [image, labels, bboxes]):
if image.dtype != tf.float32:
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
bboxes = tf_image.clip_bboxes(bboxes)
dst_image = image
dst_image, labels, bboxes = tf_image.distorted_bounding_box_crop(image, labels, bboxes)
dst_image = tf_image.distort_color(dst_image)
dst_image, bboxes = tf_image.random_flip_left_right(dst_image, bboxes)
dst_image = tf_image.resize_image(dst_image, out_shape)
num = tf.reduce_sum(tf.cast(labels, tf.int32))
image = dst_image*255.0
image = tf_image.tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
return image, labels, bboxes, num
def preprocess_for_eval(image, labels, bboxes, cord,
out_shape, data_format='NHWC',
scope='txt_preprocessing_test'):
"""Preprocess an image for evaluation.
Args:
image: A `Tensor` representing an image of arbitrary size.
labels : A Tensor inlcudes all labels
bboxes : A Tensor inlcudes cordinates of bbox in shape [N, 4]
out_shape : Image_size ,default is [300, 300]
Returns:
A preprocessed image.
"""
with tf.name_scope(scope):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
image = tf.to_float(image)
num = 0
if labels is not None:
num = tf.reduce_sum(tf.cast(labels, tf.int32))
# Add image rectangle to bboxes.
bbox_img = tf.constant([[0., 0., 1., 1.]])
if bboxes is None:
bboxes = bbox_img
else:
bboxes = tf.concat([bbox_img, bboxes], axis=0)
image = tf_image.resize_image(image, out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
image.set_shape([out_shape[0], out_shape[1], 3])
image = image / 255.
return image, labels, bboxes, cord, num
def preprocess_for_eval(image,
labels,
bboxes,
out_shape=EVAL_SIZE,
data_format='NHWC',
difficults=None,
resize=Resize.WARP_RESIZE,
scope='ssd_preprocessing_train'):
"""Preprocess an image for evaluation.
Args:
image: A `Tensor` representing an image of arbitrary size.
out_shape: Output shape after pre-processing (if resize != None)
resize: Resize strategy.
Returns:
A preprocessed image.
"""
with tf.name_scope(scope):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
image = tf.to_float(image)
image = tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
# Add image rectangle to bboxes.
bbox_img = tf.constant([[0., 0., 1., 1.]])
if bboxes is None:
bboxes = bbox_img
else:
bboxes = tf.concat([bbox_img, bboxes], axis=0)
if resize == Resize.NONE:
# No resizing...
pass
elif resize == Resize.CENTRAL_CROP:
# Central cropping of the image.
image, bboxes = tf_image.resize_image_bboxes_with_crop_or_pad(
image, bboxes, out_shape[0], out_shape[1])
elif resize == Resize.PAD_AND_RESIZE:
# Resize image first: find the correct factor...
shape = tf.shape(image)
factor = tf.minimum(
tf.to_double(1.0),
tf.minimum(tf.to_double(out_shape[0] / shape[0]),
tf.to_double(out_shape[1] / shape[1])))
resize_shape = factor * tf.to_double(shape[0:2])
resize_shape = tf.cast(tf.floor(resize_shape), tf.int32)
image = tf_image.resize_image(
image,
resize_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
# Pad to expected size.
image, bboxes = tf_image.resize_image_bboxes_with_crop_or_pad(
image, bboxes, out_shape[0], out_shape[1])
elif resize == Resize.WARP_RESIZE:
# Warp resize of the image.
image = tf_image.resize_image(
image,
out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
# Split back bounding boxes.
bbox_img = bboxes[0]
bboxes = bboxes[1:]
# Remove difficult boxes.
if difficults is not None:
mask = tf.logical_not(tf.cast(difficults, tf.bool))
labels = tf.boolean_mask(labels, mask)
bboxes = tf.boolean_mask(bboxes, mask)
# Image data format.
if data_format == 'NCHW':
image = tf.transpose(image, perm=(2, 0, 1))
return image, labels, bboxes, bbox_img
def preprocess_for_train(image,
labels,
bboxes,
out_shape,
data_format='NHWC',
scope='ssd_preprocessing_train'):
"""Preprocesses the given image for training.
Note that the actual resizing scale is sampled from
[`resize_size_min`, `resize_size_max`].
Args:
image: A `Tensor` representing an image of arbitrary size.
output_height: The height of the image after preprocessing.
output_width: The width of the image after preprocessing.
resize_side_min: The lower bound for the smallest side of the image for
aspect-preserving resizing.
resize_side_max: The upper bound for the smallest side of the image for
aspect-preserving resizing.
Returns:
A preprocessed image.
"""
fast_mode = False
with tf.name_scope(scope, 'ssd_preprocessing_train',
[image, labels, bboxes]):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
# Convert to float scaled [0, 1].
if image.dtype != tf.float32:
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
tf_summary_image(image, bboxes, 'image_with_bboxes')
# # Remove DontCare labels.
# labels, bboxes = ssd_common.tf_bboxes_filter_labels(out_label,
# labels,
# bboxes)
# Distort image and bounding boxes.
dst_image = image
dst_image, labels, bboxes, distort_bbox = \
distorted_bounding_box_crop(image, labels, bboxes,
aspect_ratio_range=CROP_RATIO_RANGE)
# Resize image to output size.
dst_image = tf_image.resize_image(
dst_image,
out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
tf_summary_image(dst_image, bboxes, 'image_shape_distorted')
# Randomly flip the image horizontally.
dst_image, bboxes = tf_image.random_flip_left_right(dst_image, bboxes)
# Randomly distort the colors. There are 4 ways to do it.
dst_image = apply_with_random_selector(
dst_image,
lambda x, ordering: distort_color(x, ordering, fast_mode),
num_cases=4)
tf_summary_image(dst_image, bboxes, 'image_color_distorted')
# Rescale to VGG input scale.
image = dst_image * 255.
image = tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
# Image data format.
if data_format == 'NCHW':
image = tf.transpose(image, perm=(2, 0, 1))
return image, labels, bboxes
def preprocess_for_train(image, labels, bboxes, xs, ys,
out_shape, data_format='NHWC',
scope='ssd_preprocessing_train', clip=True, crop_area_range=AREA_RANGE):
"""Preprocesses the given image for training.
Note that the actual resizing scale is sampled from
[`resize_size_min`, `resize_size_max`].
Args:
image: A `Tensor` representing an image of arbitrary size.
output_height: The height of the image after preprocessing.
output_width: The width of the image after preprocessing.
resize_side_min: The lower bound for the smallest side of the image for
aspect-preserving resizing.
resize_side_max: The upper bound for the smallest side of the image for
aspect-preserving resizing.
Returns:
A preprocessed image.
"""
fast_mode = False
with tf.compat.v1.name_scope(scope, 'ssd_preprocessing_train', [image, labels, bboxes]):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
orig_dtype = image.dtype
print('orig_dtype:', orig_dtype)
# Convert to float scaled [0, 1].
if image.dtype != tf.float32:
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
# tf_summary_image(image, bboxes, 'image_with_bboxes')
# Distort image and bounding boxes.
dst_image = image
dst_image, labels, bboxes,xs, ys, distort_bbox = \
distorted_bounding_box_crop(image, labels, bboxes,xs, ys,
aspect_ratio_range=CROP_RATIO_RANGE,min_object_covered=MIN_OBJECT_COVERED,area_range=crop_area_range)
# Resize image to output size.
dst_image = tf_image.resize_image(dst_image, out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
#tf_summary_image(dst_image, bboxes, 'image_shape_distorted')
# Randomly flip the image horizontally.
#bboxes and xs ys all need to random
dst_image, bboxes, xs, ys = tf_image.random_flip_left_right(dst_image, bboxes, xs, ys)
# Randomly distort the colors. There are 4 ways to do it.
dst_image = apply_with_random_selector(
dst_image,
lambda x, ordering: distort_color(x, ordering, fast_mode),
num_cases=4)
tf_summary_image(dst_image, bboxes, 'image_color_distorted')
# Rescale to VGG input scale.
image = tf.cast(tf.image.convert_image_dtype(dst_image, orig_dtype, saturate=True), dtype=tf.float32)
# image = dst_image * 255.
# image = tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
image = _mean_image_subtraction(image, [_R_MEAN, _G_MEAN, _B_MEAN] )
# Image data format.
if data_format == 'NCHW':
image = tf.transpose(a=image, perm=(2, 0, 1))
if clip:
xy_clip_min = tf.constant([0., 0., 0., 0.])
xy_clip_max = tf.constant([1., 1., 1., 1.])
bbox_img_max = tf.constant([1., 1., 1. , 1.])
bbox_img_min = tf.constant([0., 0., 0., 0.])
bboxes = tf.minimum(bboxes, bbox_img_max)
bboxes = tf.maximum(bboxes, bbox_img_min)
xs = tf.maximum(xs, xy_clip_min)
ys = tf.maximum(ys, xy_clip_min)
xs = tf.minimum(xs, xy_clip_max)
ys = tf.minimum(ys, xy_clip_max)
tf_summary_image(image, bboxes, ' image whitened')
# image = tf.Print(image, [image[0]], ' image: ', summarize=20)
# xs = tf.Print(xs, [xs, tf.shape(xs)], ' xs ', summarize=20)
# ys = tf.Print(ys, [ys, tf.shape(ys)], ' ys ', summarize=20)
# bboxes = tf.Print(bboxes, [bboxes, tf.shape(bboxes)], ' bboxes ',summarize=20)
return image, labels, bboxes, xs, ys
def preprocess_for_eval(image,
labels,
bboxes,
height,
width,
out_shape=EVAL_SIZE,
data_format='NHWC',
use_whiten=True,
difficults=None,
resize=Resize.WARP_RESIZE,
scope='ssd_preprocessing_train'):
"""Preprocess an image for evaluation.
Args:
image: A `Tensor` representing an image of arbitrary size.
labels : A Tensor inlcudes all labels
bboxes : A Tensor inlcudes cordinates of bbox in shape [N, 4]
out_shape : Image_size ,default is [300, 300]
Returns:
A preprocessed image.
"""
with tf.name_scope(scope):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
image = tf.to_float(image)
num = 0
if labels is not None:
num = tf.reduce_sum(tf.cast(labels, tf.int32))
# Add image rectangle to bboxes.
bbox_img = tf.constant([[0., 0., 1., 1.]])
if bboxes is None:
bboxes = bbox_img
else:
bboxes = tf.concat([bbox_img, bboxes], axis=0)
if resize == Resize.NONE:
# No resizing...
pass
elif resize == Resize.CENTRAL_CROP:
# Central cropping of the image.
image, bboxes = tf_image.resize_image_bboxes_with_crop_or_pad(
image, bboxes, out_shape[0], out_shape[1])
elif resize == Resize.PAD_AND_RESIZE:
# Resize image first: find the correct factor...
shape = tf.shape(image)
factor = tf.minimum(
tf.to_double(1.0),
tf.minimum(tf.to_double(out_shape[0] / shape[0]),
tf.to_double(out_shape[1] / shape[1])))
resize_shape = factor * tf.to_double(shape[0:2])
resize_shape = tf.cast(tf.floor(resize_shape), tf.int32)
image = tf_image.resize_image(
image,
resize_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
# Pad to expected size.
image, bboxes = tf_image.resize_image_bboxes_with_crop_or_pad(
image, bboxes, out_shape[0], out_shape[1])
elif resize == Resize.WARP_RESIZE:
# Warp resize of the image.
image = tf_image.resize_image(
image,
out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
# Split back bounding boxes.
bbox_img = bboxes[0]
bboxes = bboxes[1:]
# Remove difficult boxes.
if difficults is not None:
mask = tf.logical_not(tf.cast(difficults, tf.bool))
labels = tf.boolean_mask(labels, mask)
bboxes = tf.boolean_mask(bboxes, mask)
image = tf.clip_by_value(image, 0., 255.)
image = tf_image.tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
#image = image/255.
#image = tf.clip_by_value(image, 0., 255.)
#image = tf.subtract(image, 128.)
#image = tf.multiply(image, 2.0)
if data_format == 'NHWC':
image = image
else:
image = tf.transpose(image, perm=(2, 0, 1))
return image, labels, bboxes, bbox_img, num
def preprocess_for_train(image, labels, bboxes,
out_shape, data_format='NHWC',
scope='textbox_process_train'):
"""Preprocesses the given image for training.
Args:
image: A `Tensor` representing an image of arbitrary size.
labels : A Tensor inlcudes all labels
bboxes : A Tensor inlcudes cordinates of bbox in shape [N, 4]
out_shape : Image_size ,default is [300, 300]
Returns:
A preprocessed image.
"""
with tf.name_scope(scope, 'textbox_process_train', [image, labels, bboxes]):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
# Convert to float scaled [0, 1].
if image.dtype != tf.float32:
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
num = tf.reduce_sum(tf.cast(labels, tf.int32))
bboxes = tf.minimum(bboxes, 0.9999)
bboxes = tf.maximum(bboxes, 0.0001)
# Distort image and bounding boxes.
image, labels, bboxes, distort_bbox ,num= \
distorted_bounding_box_crop(image, labels, bboxes,
aspect_ratio_range=CROP_RATIO_RANGE)
# Resize image to output size.
dst_image = tf_image.resize_image(image, out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
'''
dst_image ,bboxes = \
tf_image.resize_image_bboxes_with_crop_or_pad(image, bboxes,
out_shape[0],out_shape[1])
'''
# Randomly flip the image horizontally.
dst_image, bboxes = tf_image.random_flip_left_right(dst_image, bboxes)
bbox_image = tf.image.draw_bounding_boxes(tf.expand_dims(dst_image,0), tf.expand_dims(bboxes,0))
tf.summary.image('image_with_box', bbox_image)
tf.add_to_collection('EXTRA_LOSSES', num)
dst_image = tf_image.apply_with_random_selector(
dst_image,
lambda x, ordering: tf_image.distort_color_2(x, ordering, True),
num_cases=4)
tf_image.tf_summary_image(dst_image, bboxes, 'image_color_distorted')
# Rescale to normal range
image = dst_image * 255.
dst_image.set_shape([out_shape[0], out_shape[1], 3])
image = tf_image.tf_image_whitened(image, [_R_MEAN, _G_MEAN, _B_MEAN])
#dst_image = tf.cast(dst_image,tf.float32)
return image, labels, bboxes,num
def preprocess_for_train(image,
labels,
bboxes,
out_shape,
data_format='NHWC',
use_whiten=True,
scope='textbox_process_train'):
"""Preprocesses the given image for training.
Args:
image: A `Tensor` representing an image of arbitrary size.
labels : A Tensor inlcudes all labels
bboxes : A Tensor inlcudes cordinates of bbox in shape [N, 4]
out_shape : Image_size ,default is [300, 300]
Returns:
A preprocessed image.
"""
with tf.name_scope(scope, 'textbox_process_train',
[image, labels, bboxes]):
if image.get_shape().ndims != 3:
raise ValueError('Input must be of size [height, width, C>0]')
# Convert to float scaled [0, 1].
if image.dtype != tf.float32:
image = tf.image.convert_image_dtype(image, dtype=tf.float32)
num = tf.reduce_sum(tf.cast(labels, tf.int32))
bboxes = tf.minimum(bboxes, 1.0)
bboxes = tf.maximum(bboxes, 0.0)
#image, boxes = zoom_out(image, boxes)
# Distort image and bounding boxes.
object_covered = np.random.randint(5)
min_object_covered = OBJECT_COVERED[object_covered]
image, labels, bboxes, distort_bbox ,num= \
distorted_bounding_box_crop(image, labels, bboxes,
min_object_covered=min_object_covered,
aspect_ratio_range=CROP_RATIO_RANGE)
# Resize image to output size.
dst_image = tf_image.resize_image(
image,
out_shape,
method=tf.image.ResizeMethod.BILINEAR,
align_corners=False)
dst_image, bboxes = tf_image.random_flip_left_right(dst_image, bboxes)
#dst_image.set_shape([out_shape[0], out_shape[1], 3])
#bbox_image = tf.image.draw_bounding_boxes(tf.expand_dims(dst_image,0), tf.expand_dims(bboxes,0))
#tf.summary.image('image_with_box', bbox_image)
dst_image = tf_image.apply_with_random_selector(
dst_image,
lambda x, ordering: tf_image.distort_color_2(x, ordering, False),
num_cases=4)
# Rescale to normal range
image = dst_image * 255
image.set_shape([out_shape[0], out_shape[1], 3])
if use_whiten:
image = tf_image.tf_image_whitened(image,
[_R_MEAN, _G_MEAN, _B_MEAN])
image = image / 255.0
bboxes = tf.minimum(bboxes, 1.0)
bboxes = tf.maximum(bboxes, 0.0)
return image, labels, bboxes, num
