def __init__(
self,
transform_generator=None,
batch_size=1,
group_method='ratio', # one of 'none', 'random', 'ratio'
shuffle_groups=True,
image_min_side=800,
image_max_side=1333,
transform_parameters=None,
compute_shapes=guess_shapes,
compute_anchor_targets=anchor_targets_bbox,
config=None):
self.transform_generator = transform_generator
self.batch_size = int(batch_size)
self.group_method = group_method
self.shuffle_groups = shuffle_groups
self.image_min_side = image_min_side
self.image_max_side = image_max_side
self.transform_parameters = transform_parameters or TransformParameters(
)
self.compute_shapes = compute_shapes
self.compute_anchor_targets = compute_anchor_targets
self.config = config
# Define groups
self.group_images()
# Shuffle when initializing
if self.shuffle_groups:
self.on_epoch_end()
def __init__(
self,
transform_generator=None,
batch_size=1,
group_method='ratio', # one of 'none', 'random', 'ratio'
shuffle_groups=True,
image_min_side=800,
image_max_side=1333,
transform_parameters=None,
compute_shapes=guess_shapes,
compute_anchor_targets=anchor_targets_bbox,
config=None):
self.transform_generator = transform_generator
self.batch_size = int(batch_size)
self.group_method = group_method
self.shuffle_groups = shuffle_groups
self.image_min_side = image_min_side
self.image_max_side = image_max_side
self.transform_parameters = transform_parameters or TransformParameters(
)
self.compute_shapes = compute_shapes
self.compute_anchor_targets = compute_anchor_targets
self.config = config
self.group_index = 0
self.lock = threading.Lock()
self.group_images()
def __init__(self,
csv_data_file,
csv_class_file,
base_dir=None,
transform_generator=None,
batch_size=2,
group_method='none', # one of 'none', 'random', 'ratio'
shuffle_groups=True,
image_min_side=480,
image_max_side=640,
transform_parameters=None,
compute_anchor_targets=anchor_targets_bbox,
compute_shapes=guess_shapes,
preprocess_image=preprocess_image,
config=None
):
""" Initialize a CSV data generator.
Args for csv
csv_data_file: Path to the CSV annotations file.
csv_class_file: Path to the CSV classes file.
base_dir: Directory w.r.t. where the files are to be searched (defaults to the directory containing the csv_data_file).
Args for generator
transform_generator : A generator used to randomly transform images and annotations.
batch_size : The size of the batches to generate.
group_method : Determines how images are grouped together (defaults to 'ratio', one of ('none', 'random', 'ratio')).
shuffle_groups : If True, shuffles the groups each epoch.
image_min_side : After resizing the minimum side of an image is equal to image_min_side.
image_max_side : If after resizing the maximum side is larger than image_max_side, scales down further so that the max side is equal to image_max_side.
transform_parameters : The transform parameters used for data augmentation.
compute_anchor_targets : Function handler for computing the targets of anchors for an image and its annotations.
compute_shapes : Function handler for computing the shapes of the pyramid for a given input.
preprocess_image : Function handler for preprocessing an image (scaling / normalizing) for passing through a network.
"""
self.image_names = []
self.image_data = {}
self.base_dir = base_dir
self.layout_2_group_table = {}
# take base_dir from annotations file if not explicitly specified.
if self.base_dir is None:
self.base_dir = os.path.dirname(csv_data_file)
# parse the provided class file
try:
with _open_for_csv(csv_class_file) as file:
self.classes = _read_classes(csv.reader(file, delimiter=','))
except ValueError as e:
raise_from(ValueError('invalid CSV class file: {}: {}'.format(csv_class_file, e)), None)
self.labels = {}
for key, value in self.classes.items():
self.labels[value] = key
# csv with img_path, x1, y1, x2, y2, class_name
try:
with _open_for_csv(csv_data_file) as file:
self.image_data = _read_annotations(csv.reader(file, delimiter=','), self.classes)
except ValueError as e:
raise_from(ValueError('invalid CSV annotations file: {}: {}'.format(csv_data_file, e)), None)
self.image_names = list(self.image_data.keys())
self.transform_generator = transform_generator
self.batch_size = int(batch_size)
self.group_method = group_method
self.shuffle_groups = shuffle_groups
self.image_min_side = image_min_side
self.image_max_side = image_max_side
self.transform_parameters = transform_parameters or TransformParameters()
self.compute_anchor_targets = compute_anchor_targets
self.compute_shapes = compute_shapes
self.preprocess_image = preprocess_image
if config is not None:
self.config = read_config_file(config)
else:
self.config = None
self.group_index = 0
self.lock = threading.Lock()
self.group_images()
self.layout_lookup()
from keras_retinanet.utils.transform import random_transform
from keras_retinanet.utils.image import apply_transform, TransformParameters, adjust_transform_for_image, read_image_bgr
from matplotlib.image import imsave
if __name__ == "__main__":
img = read_image_bgr('./TestFile/foo.jpg')
transform = random_transform(flip_x_chance=1, flip_y_chance=0)
print(transform)
transform_parameters = TransformParameters()
transform = adjust_transform_for_image(
transform, img, transform_parameters.relative_translation)
img_transformed = apply_transform(transform, img, transform_parameters)
imsave('./TestFile/foo_transformed.jpg', img_transformed)
exit(0)