def get_RetinaNet_model():
from keras.utils import custom_object_scope
from keras_resnet.layers import BatchNormalization
from keras_retinanet.layers import UpsampleLike, Anchors, RegressBoxes, ClipBoxes, FilterDetections
from keras_retinanet.initializers import PriorProbability
from keras_retinanet import models
from keras_retinanet.models.retinanet import retinanet_bbox
custom_objects = {
'BatchNormalization': BatchNormalization,
'UpsampleLike': UpsampleLike,
'Anchors': Anchors,
'RegressBoxes': RegressBoxes,
'PriorProbability': PriorProbability,
'ClipBoxes': ClipBoxes,
'FilterDetections': FilterDetections,
}
with custom_object_scope(custom_objects):
backbone = models.backbone('resnet50')
model = backbone.retinanet(500)
prediction_model = retinanet_bbox(model=model)
# prediction_model.load_weights("...your weights here...")
return prediction_model, custom_objects
def test_lengths(config):
"""Assert that a csv generator and tfrecords create
the same number of images in a epoch"""
created_records = tfrecords.create_tfrecords(
annotations_file="tests/output/testfile_tfrecords.csv",
class_file="tests/output/classes.csv",
image_min_side=config["image-min-side"],
backbone_model=config["backbone"],
size=100,
savedir="tests/output/")
# tfdata
tf_filenames = find_tf_filenames(path="tests/output/*.tfrecord")
# keras generator
backbone = models.backbone(config["backbone"])
generator = csv_generator.CSVGenerator(
csv_data_file="tests/output/testfile_tfrecords.csv",
csv_class_file="tests/output/classes.csv",
image_min_side=config["image-min-side"],
preprocess_image=backbone.preprocess_image,
)
fit_genertor_length = generator.size()
assert len(tf_filenames) == fit_genertor_length
def main(args=None):
from keras import backend as K
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
print('Arguments: {}'.format(args))
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# make sure keras is the minimum required version
check_keras_version()
# optionally choose specific GPU
if args.gpu:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
keras.backend.tensorflow_backend.set_session(get_session())
# create the generators
validation_generator = create_generators(args, backbone.preprocess_image)
# create the model
weights = args.weights
# default to imagenet if nothing else is specified
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
print('Creating model, this may take a second...')
model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=validation_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone)
# evaluate model
print('Evaluating model from {}, this may take a while...'.format(
args.weights))
average_precisions = evaluate(
validation_generator,
prediction_model,
iou_threshold=0.5,
score_threshold=0.05,
max_detections=300,
)
# print evaluation
present_classes = 0
precision = 0
for label, (average_precision,
num_annotations) in average_precisions.items():
print('{:.0f} instances of class'.format(num_annotations),
validation_generator.label_to_name(label),
'with average precision: {:.4f}'.format(average_precision))
if num_annotations > 0:
present_classes += 1
precision += average_precision
print('mAP: {:.4f}'.format(precision / present_classes))
def use_release(self, gpus=1):
'''Use the latest DeepForest model release from github and load model. Optionally download if release doesn't exist
Returns:
model (object): A trained keras model
gpus: number of gpus to parallelize, default to 1
'''
#Download latest model from github release
release_tag, self.weights = utilities.use_release()
#load saved model and tag release
self.__release_version__ = release_tag
print("Loading pre-built model: {}".format(release_tag))
if gpus == 1:
with warnings.catch_warnings():
#Suppress compilte warning, not relevant here
warnings.filterwarnings("ignore", category=UserWarning)
self.model = utilities.read_model(self.weights, self.config)
#Convert model
self.prediction_model = convert_model(self.model)
elif gpus > 1:
backbone = models.backbone(self.config["backbone"])
n_classes = len(self.labels.keys())
self.model, self.training_model, self.prediction_model = create_models(
backbone.retinanet,
num_classes=n_classes,
weights=self.weights,
multi_gpu=gpus)
#add to config
self.config["weights"] = self.weights
def __init__(self,
backbone="resnet50",
random_transform=False,
batch_size=2,
step=100,
epoch=40,
anchor_settings=None):
""" Initializer
Args:
backbone
string indicate which backbone network to use (only for ones defined in
keras-retinanet)
random_transform
boolean indicate whether image is randomly transformed when training
batch_size
number of images sent for each step during training
step
number of steps(batches) to run for each epoch during training
epoch
number of epochs for this training
anchor_settings
a list of lists indicate anchor box parameters. please reference
keras-retinanet's document for how to set up anchor box parameters
"""
check_keras_version()
self._backbone_name = backbone
self._backbone = models.backbone(backbone)
self._img_preprocessor = self._backbone.preprocess_image
self._batch_size = batch_size
self._step = step
self._epoch = epoch
if random_transform:
self._transform_generator = random_transform_generator(
min_rotation=-0.1,
max_rotation=0.1,
min_translation=(-0.1, -0.1),
max_translation=(0.1, 0.1),
min_shear=-0.1,
max_shear=0.1,
min_scaling=(0.9, 0.9),
max_scaling=(1.1, 1.1),
flip_x_chance=0.5,
flip_y_chance=0.5,
)
else:
self._transform_generator = random_transform_generator(
flip_x_chance=0.5)
self._common_args = {
'batch_size': batch_size,
'preprocess_image': self._img_preprocessor
}
if anchor_settings:
self._anchor_params = AnchorParameters(*anchor_settings)
else:
self._anchor_params = AnchorParameters.default
def _load_model_with_nms(self, test_args):
""" This is mostly copied fomr retinanet.py """
backbone_name = test_args.get('DETECTOR', 'backbone')
print(backbone_name)
print(test_args.get('DETECTOR', 'detector_model_path'))
model = keras.models.load_model(
str(test_args.get('DETECTOR', 'detector_model_path')),
custom_objects=backbone(backbone_name).custom_objects)
# compute the anchors
features = [
model.get_layer(name).output
for name in ['P3', 'P4', 'P5', 'P6', 'P7']
]
anchors = build_anchors(AnchorParameters.default, features)
# we expect the anchors, regression and classification values as first
# output
print(len(model.outputs))
regression = model.outputs[0]
classification = model.outputs[1]
print(classification.shape[1])
print(regression.shape)
# "other" can be any additional output from custom submodels,
# by default this will be []
other = model.outputs[2:]
# apply predicted regression to anchors
boxes = layers.RegressBoxes(name='boxes')([anchors, regression])
boxes = layers.ClipBoxes(name='clipped_boxes')(
[model.inputs[0], boxes])
# filter detections (apply NMS / score threshold / select top-k)
#detections = layers.FilterDetections(
# nms=True,
# name='filtered_detections',
# nms_threshold = test_args.getfloat('DETECTOR','nms_threshold'),
# score_threshold = test_args.getfloat('DETECTOR','det_threshold'),
# max_detections = test_args.getint('DETECTOR', 'max_detections')
# )([boxes, classification] + other)
detections = layers.filter_detections.filter_detections(
boxes=boxes,
classification=classification,
other=other,
nms=True,
nms_threshold=test_args.getfloat('DETECTOR', 'nms_threshold'),
score_threshold=test_args.getfloat('DETECTOR', 'det_threshold'),
max_detections=test_args.getint('DETECTOR', 'max_detections'))
outputs = detections
# construct the model
return keras.models.Model(inputs=model.inputs,
outputs=outputs,
name='retinanet-bbox')
def main(args=None):
args = parse_args(args)
action = args.action
backbone = models.backbone("resnet50")
modelpath = args.model_path
cocopath = args.coco_path
generator = create_generator(args, backbone.preprocess_image)
quantize_retinanet(modelpath, cocopath, action)
evaluate(generator, action, args.score_threshold)
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
# optionally choose specific GPU
# if args.gpu:
# os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
with tf.device('/device:GPU:3'):
config = tf.ConfigProto(allow_soft_placement=True, log_device_placement=False, device_count={'CPU': 1, 'GPU': 1})
config.gpu_options.allow_growth = True
session = tf.Session(config=config)
keras.backend.set_session(session)
# create the generators
val_generator = create_val_generator(args)
weights = args.weights
# default to imagenet if nothing else is specified
backbone = models.backbone(args.backbone)
print('Creating model, this may take a second...')
# model = model_with_weights(backbone(val_generator.num_classes(), num_anchors=None, modifier=freeze_model), weights=weights, skip_mismatch=True)
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=val_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
lr=args.lr,
config=args.config
)
# print model summary
print(model.summary())
evaluation = Evaluate(val_generator, weighted_average=args.weighted_average, eval_batch_size=args.val_batch_size)
evaluation = RedirectModel(evaluation, prediction_model)
generator = val_generator
iou_threshold = 0.5
score_threshold = 0.05
max_detections = 100
save_path = None
weighted_average = False
verbose = True
eval_batch_size = 16
from keras_retinanet.utils.sail_eval_data_generator import SAIL_EvalDataGenerator
eval_batch_datagen = SAIL_EvalDataGenerator(generator, batch_size=eval_batch_size)
boxes, scores, labels = model.predict_generator(eval_batch_datagen, steps=len(eval_batch_datagen), verbose=True,
workers=0, use_multiprocessing=False)[:3]
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# optionally load config parameters
if args.config:
args.config = read_config_file(args.config)
# create the generators
train_generator = create_generators(args, backbone.preprocess_image)
# create the model
if args.snapshot is not None:
print('Loading model')
model = models.load_model(args.snapshot, backbone_name=args.backbone)
training_model = model
anchor_params = None
if args.config and 'anchor_parameters' in args.config:
anchor_params = parse_anchor_parameters(args.config)
prediction_model = retinanet_bbox(model=model,
anchor_params=anchor_params)
else:
weights = args.weights
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
print('Creating model')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
lr=args.lr,
config=args.config)
# this lets the generator compute backbone layer shapes using the actual backbone model
if 'densenet' in args.backbone:
train_generator.compute_shapes = make_shapes_callback(model)
# create the callbacks
callbacks = create_callbacks(model, training_model, prediction_model, args)
# start training
return training_model.fit_generator(
generator=train_generator,
steps_per_epoch=train_generator.size() // args.batch_size,
epochs=args.epochs,
verbose=1,
callbacks=callbacks)
def load_retina(weights: str, num_classes: int,
anchors_wrap: AnchorParametersWrap,
backbone_name: str) -> Model:
""" Loads retinanet with weights. """
ret = models.backbone(backbone_name=backbone_name).retinanet(
num_classes=num_classes, num_anchors=anchors_wrap.num_anchors)
ret.load_weights(weights)
ret = models.convert_model(model=ret,
nms=True,
class_specific_filter=True,
anchor_params=anchors_wrap.anchors)
return ret
def get_joint_detection_model(model_path, model_type):
"""
Input -> Model path for the object detection model
Model type-> Foot or Hand
Output -> Inference model for getting the predictions on test images
"""
# config_file_path = '/usr/local/bin/config'
if model_type == 'Foot_detection':
# with open('/usr/local/bin/src/config.ini','w') as f:
# f.write('[anchor_parameters]\nsizes = 32 64 128 256 512 1024\nstrides = 8 16 32 64 128 256\nratios = 1.2 1.5 2 2.5 3\nscales =1 1.5 2\n')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone('resnet50').retinanet,
num_classes=5,
weights=None,
multi_gpu=False,
freeze_backbone=True,
lr=1e-3,
config=read_config_file('/usr/local/bin/Config files/config_foot.ini'))
training_model.load_weights(model_path)
infer_model = convert_model(training_model, anchor_params = parse_anchor_parameters(read_config_file('/usr/local/bin/Config files/config_foot.ini')))
elif model_type == 'Hand_detection':
# with open('/usr/local/bin/src/config.ini','w') as f:
# f.write('[anchor_parameters]\nsizes = 32 64 128 256 512 1024\nstrides = 8 16 32 64 128 256\nratios = 1 1.5 2 2.5 3\nscales = 1 1.2 1.6\n')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone('resnet50').retinanet,
num_classes=6,
weights=None,
multi_gpu=False,
freeze_backbone=True,
lr=1e-3,
config=read_config_file('/usr/local/bin/Config files/config_hand.ini'))
training_model.load_weights(model_path)
infer_model = convert_model(training_model, anchor_params = parse_anchor_parameters(read_config_file('/usr/local/bin/Config files/config_hand.ini')))
return infer_model
def retinanet_train(tiny=False):
labels_path = bdd100k_labels_path
val_json = labels_path + 'bdd100k_labels_images_val.json'
train_json = labels_path + 'bdd100k_labels_images_train.json'
val_annot = labels_path + 'val_annotations.csv'
train_annot = labels_path + 'train_annotations.csv'
num_data = 70000
classes = bu.annotate(val_json, val_annot, labels_path, bdd100k_val_path)
cl_map_path = bu.class_mapping(classes,
output_csv=labels_path +
'class_mapping.csv')
bu.annotate(train_json, train_annot, bdd100k_labels_path,
bdd100k_train_path)
# Hyper-parameters
batch_size = 1
steps_per_epoch = np.ceil(num_data / batch_size)
for m in models:
print('Generating %s backbone...' % m)
backbone = kr_models.backbone(m)
weights = backbone.download_imagenet()
print('Creating generators...')
tr_gen, val_gen = mt.create_generators(
train_annotations=train_annot,
val_annotations=val_annot,
class_mapping=cl_map_path,
base_dir='',
preprocess_image=backbone.preprocess_image,
batch_size=batch_size)
print('Creating models...')
model, training_model, prediction_model = kr_train.create_models(
backbone.retinanet, tr_gen.num_classes(), weights)
print('Creating callbacks...')
callbacks = mt.create_callbacks(model,
batch_size,
'test',
tensorboard_dir=log_path)
print('Training...')
training_model.fit_generator(
generator=tr_gen,
steps_per_epoch=steps_per_epoch, # 10000,
epochs=2,
verbose=1,
callbacks=callbacks,
workers=4, # 1
use_multiprocessing=True, # False,
max_queue_size=10,
validation_data=val_gen)
def create_model():
model_path = Path(
'/data/students_home/fschipani/thesis/keras-retinanet/snapshots/resnet50_coco_best_v2.1.0.h5'
)
model = models.backbone('resnet50').retinanet(
num_classes=train_generator.num_classes())
model.load_weights(model_path, by_name=True, skip_mismatch=True)
model.compile(loss={
'regression': smooth_l1(),
'classification': focal()
},
optimizer=keras.optimizers.adam(lr=1e-5, clipnorm=0.001))
return model
def load_training_model(num_classes, args):
backbone = models.backbone(args.backbone)
weights = backbone.download_imagenet()
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=num_classes,
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
lr=args.lr,
config=args.config)
return training_model
def retinanet_tiny_train():
labels_path = bdd100k_labels_path
val_json = labels_path + 'bdd100k_labels_images_val.json'
num_data = 7000
batch_size = 1
steps_per_epoch = np.ceil(num_data / batch_size)
train_annot, val_annot = bu.annotate_tiny(val_json,
labels_path,
bdd100k_val_path,
overwrite=True)
cl_map_path = bu.class_mapping(input_json=val_json,
output_csv=labels_path +
'class_mapping.csv')
for m in models:
print('Generating %s backbone...' % m)
backbone = kr_models.backbone(m)
weights = backbone.download_imagenet()
print('Creating generators...')
tr_gen, val_gen = bu.create_generators(
train_annotations=train_annot,
val_annotations=val_annot,
class_mapping=cl_map_path,
base_dir='',
preprocess_image=backbone.preprocess_image,
batch_size=batch_size)
print('Creating models...')
model, training_model, prediction_model = kr_train.create_models(
backbone.retinanet, tr_gen.num_classes(), weights)
print('Creating callbacks...')
callbacks = bu.create_callbacks(model,
batch_size,
snapshots_path=retinanet_h5_path,
tensorboard_dir=log_path,
backbone=m,
dataset_type='bdd10k')
print('Training...')
training_model.fit_generator(
generator=tr_gen,
steps_per_epoch=steps_per_epoch, # 10000,
epochs=50,
verbose=1,
callbacks=callbacks,
workers=1, # 1
use_multiprocessing=False, # False,
max_queue_size=10,
validation_data=val_gen)
def _initilize_model(num_classes):
"""Initilze our model to train with
"""
# Suggested to initialize model on cpu before turning into a
# multi_gpu model to save gpu memory
with tf.device('/cpu:0'):
model = models.backbone('resnet50').retinanet(num_classes=num_classes)
model.load_weights(config.WEIGHTS_PATH,
by_name=True, skip_mismatch=True)
gpus = len([i for i in device_lib.list_local_devices()
if i.device_type == 'GPU'])
if gpus > 1:
return model, multi_gpu_model(model, gpus=gpus)
return model, model
def __init__(self, weights=None, saved_model=None):
self.weights = weights
self.saved_model = saved_model
# Read config file - if a config file exists in local dir use it,
# if not use installed.
if os.path.exists("deepforest_config.yml"):
config_path = "deepforest_config.yml"
else:
try:
config_path = get_data("deepforest_config.yml")
except Exception as e:
raise ValueError(
"No deepforest_config.yml found either in local "
"directory or in installed package location. {}".format(e))
print("Reading config file: {}".format(config_path))
self.config = utilities.read_config(config_path)
# Create a label dict, defaults to "Tree"
self.read_classes()
# release version id to flag if release is being used
self.__release_version__ = None
# Load saved model if needed
if self.saved_model:
print("Loading saved model")
# Capture user warning, not relevant here
with warnings.catch_warnings():
warnings.filterwarnings("ignore", category=UserWarning)
self.model = models.load_model(saved_model)
self.prediction_model = convert_model(self.model)
elif self.weights:
print("Creating model from weights")
backbone = models.backbone(self.config["backbone"])
self.model, self.training_model, self.prediction_model = create_models(
backbone.retinanet, num_classes=1, weights=self.weights)
else:
print(
"A blank deepforest object created. "
"To perform prediction, either train or load an existing model."
)
self.model = None
def run(args=None):
args = parse_args(args)
backbone = models.backbone(args.backbone)
if args.config:
args.config = read_config_file(args.config)
_, generator = create_generators(args, backbone.preprocess_image)
weights = os.path.join(args.snapshot_path, 'checkpoint.h5')
_, _, model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=10,
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
lr=args.lr,
config=args.config
)
infer(generator, model)
def test_equivalence(config, setup_create_tensors):
# unpack created tensors
tf_inputs, tf_targets = setup_create_tensors
# the image going in to tensorflow should be equivalent
# to the image from the fit_generator
backbone = models.backbone(config["backbone"])
# CSV generator
generator = csv_generator.CSVGenerator(
csv_data_file="tests/output/testfile_tfrecords.csv",
csv_class_file="tests/data/classes.csv",
image_min_side=config["image-min-side"],
preprocess_image=backbone.preprocess_image,
)
# find file in randomize generator group
first_file = generator.groups[0][0]
gen_filename = os.path.join(generator.base_dir, generator.image_names[first_file])
original_image = generator.load_image(first_file)
inputs, targets = generator.__getitem__(0)
image = inputs[0, ...]
targets = targets[0][0, ...]
with tf.Session() as sess:
# seek the randomized image to match
tf_inputs, tf_targets = sess.run([tf_inputs, tf_targets])
# assert filename is the same as generator
# assert gen_filename == filename
# tf_image = tf_image[0,...]
tf_inputs = tf_inputs[0, ...]
tf_targets = tf_targets[0][0, ...]
# Same shape
# assert tf_image.shape == image.shape
assert tf_inputs.shape == image.shape
assert tf_targets.shape == targets.shape
def test_multigpu_training():
experiment = Experiment(api_key="ypQZhYfs3nSyKzOfz13iuJpj2",
project_name='deeplidar',
log_code=True)
DeepForest_config = config.load_config(dir="..")
DeepForest_config["save_image_path"] = "../snapshots/"
data = generators.load_retraining_data(DeepForest_config)
train_generator, validation_generator = generators.create_h5_generators(
data, DeepForest_config=DeepForest_config)
#imagenet pretraining weights
backbone = models.backbone(DeepForest_config["backbone"])
weights = backbone.download_imagenet()
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
multi_gpu=DeepForest_config["num_GPUs"],
freeze_backbone=False,
nms_threshold=DeepForest_config["nms_threshold"],
input_channels=DeepForest_config["input_channels"])
#start training
history = training_model.fit_generator(
generator=train_generator,
steps_per_epoch=train_generator.size() /
DeepForest_config["batch_size"],
epochs=DeepForest_config["epochs"],
verbose=2,
shuffle=False,
workers=DeepForest_config["workers"],
use_multiprocessing=DeepForest_config["use_multiprocessing"],
max_queue_size=DeepForest_config["max_queue_size"],
experiment=experiment)
evaluation = True
# Хэрэв сургалтаа үргэлжлүүлэх бол snapshot-ын утганд сүүлийн сургалтын файлыг зааж өгнө
snapshot = './snapshots/resnet50_csv_03.h5'
# Snapshot хадгалах эсэх, хадгалах хавтас
snapshots = True
snapshot_path = './snapshots'
tensorboard_dir = './logs'
weights = None
imagenet_weights = True
batch_size = 1
image_min_side = 225
image_max_side = 300
backbone_model = models.backbone('resnet50')
freeze_backbone = False
common_args = {
'batch_size': batch_size,
'image_min_side': image_min_side,
'image_max_side': image_max_side,
'preprocess_image': backbone_model.preprocess_image,
}
train_generator = CSVGenerator(annotations_path, classes_path, **common_args)
if validations_path != '':
validation_generator = CSVGenerator(validations_path, classes_path,
**common_args)
else:
validation_generator = None
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# make sure keras is the minimum required version
check_keras_version()
# optionally choose specific GPU
if args.gpu:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
keras.backend.set_session(get_session())
# optionally load config parameters
if args.config:
args.config = read_config_file(args.config)
# create the generators
train_generator, validation_generator = create_generators(
args, backbone.preprocess_image)
# create the model
if args.snapshot is not None:
print('Loading model, this may take a second...')
model = models.load_model(args.snapshot, backbone_name=args.backbone)
training_model = model
anchor_params = None
if args.config and 'anchor_parameters' in args.config:
anchor_params = parse_anchor_parameters(args.config)
prediction_model = retinanet_bbox(model=model,
anchor_params=anchor_params)
else:
weights = args.weights
# default to imagenet if nothing else is specified
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
print('Creating model, this may take a second...')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
lr=args.lr,
config=args.config)
# print model summary
print(model.summary())
# this lets the generator compute backbone layer shapes using the actual backbone model
# if 'vgg' in args.backbone or 'densenet' in args.backbone:
# train_generator.compute_shapes = make_shapes_callback(model)
# if validation_generator:
# validation_generator.compute_shapes = train_generator.compute_shapes
# create the callbacks
callbacks = create_callbacks(model,
training_model,
prediction_model,
validation_generator,
args,
eval_batch_size=args.val_batch_size)
# Use multiprocessing if workers > 0
if args.workers > 0:
use_multiprocessing = True
else:
use_multiprocessing = False
# start training
training_model.fit_generator(generator=train_generator,
steps_per_epoch=args.steps,
epochs=args.epochs,
verbose=1,
callbacks=callbacks,
workers=args.workers,
use_multiprocessing=use_multiprocessing,
max_queue_size=args.max_queue_size)
def test_equivalence(config, setup_create_tensors):
#unpack created tensors
tf_inputs, tf_targets = setup_create_tensors
#the image going in to tensorflow should be equivalent to the image from the fit_generator
backbone = models.backbone(config["backbone"])
#CSV generator
generator = csv_generator.CSVGenerator(
csv_data_file="tests/data/testfile_tfrecords.csv",
csv_class_file="tests/data/classes.csv",
image_min_side=config["image-min-side"],
preprocess_image=backbone.preprocess_image,
)
#find file in randomize generator group
first_file = generator.groups[0][0]
gen_filename = os.path.join(generator.base_dir, generator.image_names[first_file])
original_image = generator.load_image(first_file)
inputs, targets = generator.__getitem__(0)
image = inputs[0,...]
targets = targets[0][0,...]
with tf.Session() as sess:
#seek the randomized image to match
tf_inputs, tf_targets = sess.run([tf_inputs,tf_targets])
#assert filename is the same as generator
#assert gen_filename == filename
#tf_image = tf_image[0,...]
tf_inputs = tf_inputs[0,...]
tf_targets = tf_targets[0][0,...]
#Same shape
#assert tf_image.shape == image.shape
assert tf_inputs.shape == image.shape
assert tf_targets.shape == targets.shape
#Same values, slightly duplicitious with above, but useful for debugging.
#Saved array is the same as generator image
#np.testing.assert_array_equal(image, tf_image)
#Loaded array is the same as generator, this is not true currently, the opencv and the tensorflow interpolation method is slightly different, waiting for tf. 2.0
#np.testing.assert_array_equal(tf_loaded, tf_image)
##Useful for debug to plot
#fig = plt.figure()
#ax1 = fig.add_subplot(1,4,1)
#ax1.title.set_text('Fit Gen Original')
#plt.imshow(original_image[...,::-1])
#ax1 = fig.add_subplot(1,4,2)
#ax1.title.set_text('Fit Generator')
#plt.imshow(image)
#ax2 = fig.add_subplot(1,4,3)
#ax2 = fig.add_subplot(1,4,4)
#ax2.title.set_text('Loaded Image')
#plt.imshow(tf_inputs)
#plt.show()
#Check for bad file types
#@pytest.fixture()
#def bad_annotations():
#annotations = utilities.xml_to_annotations(get_data("OSBS_029.xml"))
#f = "tests/data/testfile_error_deepforest.csv"
#annotations.to_csv(f,index=False,header=False)
#return f
#def test_tfdataset_error(bad_annotations):
#with pytest.raises(ValueError):
#records_created = tfrecords.create_tfrecords(annotations_file=bad_annotations, class_file=get_data("classes.csv"), image_min_side=800, backbone_model="resnet50", size=100, savedir="tests/data/")
'regression': losses.smooth_l1(),
'classification': losses.focal()
},
optimizer=keras.optimizers.adam(lr=lr,
clipnorm=0.001))
return model, training_model, prediction_model
if __name__ == '__main__':
smin, smax = 618, 800
keras.backend.tensorflow_backend.set_session(get_session())
model_path = os.path.join('snapshots', '7', 'resnet18_csv_19.h5')
backbone = models.backbone('resnet18')
labels_to_names = {
0: 'goodhelmet',
1: 'LP',
2: 'goodshoes',
3: 'badshoes',
4: 'badhelmet',
5: 'person'
}
# labels_to_names = {0: 'person', 1: 'helmet', 2: 'LP', 3: 'goodshoes', 4: 'badshoes'}
# labels_to_names = {0: 'person', 1: 'LP', 2: 'badshoes', 3: 'unsafe_hat', 4: 'Boot', 5: 'helmet',
# 6: 'goodshoes'}
main_model, training_model, prediction_model = create_models(
backbone.retinanet, len(labels_to_names))
main_model.load_weights(model_path)
def main(args=None):
global config
from keras import backend as K
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
print('Arguments: {}'.format(args))
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# make sure keras is the minimum required version
check_keras_version()
# optionally choose specific GPU
if args.gpu:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
keras.backend.tensorflow_backend.set_session(get_session())
# create the generators
train_generator, validation_generator = create_generators(
args, backbone.preprocess_image)
# create the model
if args.snapshot is not None:
print('Loading model, this may take a second...')
model = models.load_model(args.snapshot, backbone_name=args.backbone)
training_model = model
anchor_params = None
if 'anchor_parameters' in config:
anchor_params = parse_anchor_parameters(config)
prediction_model = retinanet_bbox(model=model,
anchor_params=anchor_params)
else:
weights = args.weights
# default to imagenet if nothing else is specified
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
print('Creating model, this may take a second...')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
args=args,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
config=config)
# print model summary
print(model.summary())
print('Learning rate: {}'.format(K.get_value(model.optimizer.lr)))
if args.lr > 0.0:
K.set_value(model.optimizer.lr, args.lr)
print('Updated learning rate: {}'.format(
K.get_value(model.optimizer.lr)))
# this lets the generator compute backbone layer shapes using the actual backbone model
if 'vgg' in args.backbone or 'densenet' in args.backbone:
train_generator.compute_shapes = make_shapes_callback(model)
if validation_generator:
validation_generator.compute_shapes = train_generator.compute_shapes
# create the callbacks
callbacks = create_callbacks(
model,
training_model,
prediction_model,
validation_generator,
args,
)
init_epoch = 0
try:
if args.snapshot:
init_epoch = int(args.snapshot.split("_")[-2])
except:
pass
# init_epoch = 6
print('Init epoch: {}'.format(init_epoch))
# start training
training_model.fit_generator(
generator=train_generator,
steps_per_epoch=args.steps,
epochs=args.epochs,
verbose=1,
callbacks=callbacks,
initial_epoch=init_epoch,
)
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# make sure keras is the minimum required version
check_keras_version()
# optionally choose specific GPU
if args.gpu:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
keras.backend.tensorflow_backend.set_session(get_session())
# create the generators
train_generator, validation_generator = create_generators(
args, backbone.preprocess_image)
# create the model
if args.snapshot is not None:
print('Loading model, this may take a second...')
model = models.load_model(args.snapshot, backbone_name=args.backbone)
training_model = model
prediction_model = retinanet_bbox(model=model)
else:
weights = args.weights
# default to imagenet if nothing else is specified
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
else:
print("Don't load weight!")
print('Creating model, this may take a second...')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone)
# print model summary
print(model.summary())
# create the callbacks
callbacks = create_callbacks(
model,
training_model,
prediction_model,
validation_generator,
args,
)
# start training
training_model.fit_generator(generator=train_generator,
steps_per_epoch=args.steps,
epochs=args.epochs,
verbose=1,
callbacks=callbacks,
validation_data=validation_generator,
validation_steps=20)
def main(args=None, data=None, DeepForest_config=None, experiment=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# make sure keras is the minimum required version
check_keras_version()
# optionally choose specific GPU
if args.gpu:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
keras.backend.tensorflow_backend.set_session(get_session())
# create the generators
train_generator, validation_generator = create_generators(
args, data, DeepForest_config=DeepForest_config)
# create the model
if args.snapshot is not None:
print('Loading model, this may take a secondkeras-retinanet.\n')
model = models.load_model(args.snapshot, backbone_name=args.backbone)
training_model = model
prediction_model = retinanet_bbox(
model=model, nms_threshold=DeepForest_config["nms_threshold"])
else:
weights = args.weights
# default to imagenet if nothing else is specified
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
print('Creating model, this may take a secondkeras-retinanet .')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone,
nms_threshold=DeepForest_config["nms_threshold"])
# print model summary
#print(model.summary())
# this lets the generator compute backbone layer shapes using the actual backbone model
if 'vgg' in args.backbone or 'densenet' in args.backbone:
compute_anchor_targets = functools.partial(
anchor_targets_bbox, shapes_callback=make_shapes_callback(model))
train_generator.compute_anchor_targets = compute_anchor_targets
if validation_generator is not None:
validation_generator.compute_anchor_targets = compute_anchor_targets
matched = []
for entry in validation_generator.image_data.values():
test = entry in train_generator.image_data.values()
matched.append(test)
if sum(matched) > 0:
raise Exception(
"%.2f percent of validation windows are in training data" %
(100 * sum(matched) / train_generator.size()))
else:
print("Test passed: No overlapping data in training and validation")
#start training
cp.enable()
training_model.fit_generator(generator=train_generator,
steps_per_epoch=train_generator.size() /
DeepForest_config["batch_size"],
epochs=args.epochs,
verbose=1,
shuffle=False,
callbacks=None,
workers=DeepForest_config["workers"],
use_multiprocessing=True)
cp.disable()
'classification': losses.focal()
},
optimizer=keras.optimizers.adam(lr=lr,
clipnorm=0.001))
return model, training_model, prediction_model
if __name__ == '__main__':
smin, smax = 1200, 1600
keras.backend.tensorflow_backend.set_session(get_session())
model_path = os.path.join('snapshots', '5', 'resnet50_csv_05.h5')
backbone = models.backbone('resnet50')
labels_to_names = {1: 'mif', 0: 'ov'}
main_model, training_model, prediction_model = create_models(
backbone.retinanet, len(labels_to_names))
main_model.load_weights(model_path)
model = prediction_model
path = '/media/palm/data/MicroAlgae/images'
pad = 0
ls = [
s.split(',')[0]
for s in open('datastuff/algea/val_annotations').read().split('\n')
]
found = []
# while True:
# p = os.path.join(path, np.random.choice(os.listdir(path)))
def load_pb_model(pb_model_path):
start = time.time()
pb_model = keras.models.load_model(pb_model_path, custom_objects=models.backbone('resnet50').custom_objects)
print("Load .pb model - processing time: ", time.time() - start)
return pb_model
def main(args=None):
# parse arguments
if args is None:
args = sys.argv[1:]
args = parse_args(args)
# create object that stores backbone information
backbone = models.backbone(args.backbone)
# make sure keras is the minimum required version
check_keras_version()
# optionally choose specific GPU
if args.gpu:
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
keras.backend.tensorflow_backend.set_session(get_session())
# create the generators
train_generator, validation_generator = create_generators(args, backbone.preprocess_image)
# create the model
if args.snapshot is not None:
print('Loading model, this may take a second...')
model = models.load_model(args.snapshot, backbone_name=args.backbone)
training_model = model
prediction_model = retinanet_bbox(model=model)
else:
weights = args.weights
# default to imagenet if nothing else is specified
if weights is None and args.imagenet_weights:
weights = backbone.download_imagenet()
print('Creating model, this may take a second...')
model, training_model, prediction_model = create_models(
backbone_retinanet=backbone.retinanet,
num_classes=train_generator.num_classes(),
weights=weights,
multi_gpu=args.multi_gpu,
freeze_backbone=args.freeze_backbone
)
# print model summary
print(model.summary())
# this lets the generator compute backbone layer shapes using the actual backbone model
if 'vgg' in args.backbone or 'densenet' in args.backbone:
train_generator.compute_shapes = make_shapes_callback(model)
if validation_generator:
validation_generator.compute_shapes = train_generator.compute_shapes
# create the callbacks
callbacks = create_callbacks(
model,
training_model,
prediction_model,
validation_generator,
args,
)
# start training
history=training_model.fit_generator(
generator=train_generator,
steps_per_epoch=args.steps,
epochs=args.epochs,
verbose=1,
callbacks=callbacks,
)
timestr=time.strftime("%Y-%m-%d-%H%M")
history_path=os.path.join(
args.snapshot_path,
'{timestr}_{backbone}.csv'.format(timestr=timestr,backbone=args.backbone, dataset_type=args.dataset_type)
)
pd.DataFrame(history.history).to_csv(history_path)
