def get_datagen(self, batch_size, mode='train', view=False):
if mode == 'train':
augmentors = self.get_train_augmentors(self.train_input_shape,
self.train_mask_shape, view)
data_files = get_files(self.train_dir, self.data_ext)
data_generator = loader.train_generator
nr_procs = self.nr_procs_train
else:
augmentors = self.get_valid_augmentors(self.infer_input_shape,
self.infer_mask_shape, view)
data_files = get_files(self.valid_dir, self.data_ext)
data_generator = loader.valid_generator
nr_procs = self.nr_procs_valid
# set nr_proc=1 for viewing to ensure clean ctrl-z
nr_procs = 1 if view else nr_procs
dataset = loader.DatasetSerial(data_files)
datagen = data_generator(dataset,
shape_aug=augmentors[0],
input_aug=augmentors[1],
label_aug=augmentors[2],
batch_size=batch_size,
nr_procs=nr_procs)
return datagen
def get_datagen(self, opt, mode='train', view=False):
if mode == 'train':
augmentors = self.get_train_augmentors(
self.train_input_shape, self.train_mask_shape,
opt['model_flags']['type_classification'], view)
data_files = get_files(opt['train_dir'], self.data_ext)
data_generator = loader.train_generator
nr_procs = self.nr_procs_train
batch_size = opt['train_batch_size']
else:
augmentors = self.get_valid_augmentors(self.infer_input_shape,
self.infer_mask_shape, view)
data_files = get_files(opt['valid_dir'], self.data_ext)
data_generator = loader.valid_generator
nr_procs = self.nr_procs_valid
batch_size = opt['infer_batch_size']
print('TRAINER:GET_DATAGEN - Batch Size: ', batch_size)
# set nr_proc=1 for viewing to ensure clean ctrl-z
nr_procs = 1 if view else nr_procs
dataset = loader.DatasetSerial(data_files)
datagen = data_generator(dataset,
shape_aug=augmentors[0],
input_aug=augmentors[1],
label_aug=augmentors[2],
batch_size=batch_size,
nr_procs=nr_procs)
return datagen
def get_datagen(self, batch_size, mode="train", view=False):
if mode == "train":
augmentors = self.get_train_augmentors(self.train_input_shape,
self.train_output_shape,
view)
data_files = get_files(self.train_dir, self.data_ext)
# Different data generators for segmentation and classification
if self.model_mode == "seg_gland" or self.model_mode == "seg_nuc":
data_generator = loader.train_generator_seg
else:
data_generator = loader.train_generator_class
nr_procs = self.nr_procs_train
else:
augmentors = self.get_valid_augmentors(self.train_input_shape,
self.train_output_shape,
view)
# Different data generators for segmentation and classification
data_files = get_files(self.valid_dir, self.data_ext)
if self.model_mode == "seg_gland" or self.model_mode == "seg_nuc":
data_generator = loader.valid_generator_seg
else:
data_generator = loader.valid_generator_class
nr_procs = self.nr_procs_valid
# set nr_proc=1 for viewing to ensure clean ctrl-z
nr_procs = 1 if view else nr_procs
dataset = loader.DatasetSerial(data_files)
if self.model_mode == "seg_gland" or self.model_mode == "seg_nuc":
datagen = data_generator(
dataset,
shape_aug=augmentors[0],
input_aug=augmentors[1],
label_aug=augmentors[2],
batch_size=batch_size,
nr_procs=nr_procs,
)
else:
datagen = data_generator(
dataset,
shape_aug=augmentors[0],
input_aug=augmentors[1],
batch_size=batch_size,
nr_procs=nr_procs,
)
return datagen