def generate_training_data():
file_name_save = os.path.join(NPZ_DIR, PREFIX, DATA_FILE)
num_of_features = 2 # Specify the number of feature masks that are present
window_size = (30, 30) # Size of window around pixel
training_direcs = ['set1', 'set2']
channel_names = ['Phase']
raw_image_direc = 'raw'
annotation_direc = 'annotated'
# Create the training data
make_training_data(
direc_name=os.path.join(DATA_DIR, PREFIX),
dimensionality=2,
max_training_examples=1e6, # Define maximum number of training examples
window_size_x=window_size[0],
window_size_y=window_size[1],
border_mode=BORDER_MODE,
file_name_save=file_name_save,
training_direcs=training_direcs,
channel_names=channel_names,
num_of_features=num_of_features,
raw_image_direc=raw_image_direc,
annotation_direc=annotation_direc,
reshape_size=RESHAPE_SIZE if RESIZE else None,
edge_feature=[1, 0, 0], # Specify which feature is the edge feature,
dilation_radius=1,
output_mode=DATA_OUTPUT_MODE,
display=False,
verbose=True)
direc_name = '/data/training_data/nuclei_broad/'
file_name_save = os.path.join('/data/training_data_npz/nuclei_broad/',
'nuclei_broad_valid_conv_61x61.npz')
training_direcs = ['set1', 'set2', 'set3', 'set4', 'set5']
channel_names = ['nuclear']
# Specify the number of feature masks that are present
num_of_features = 2
# Specify which feature is the edge feature
edge_feature = [1, 0, 0]
# Create the training data
make_training_data(max_training_examples=max_training_examples,
window_size_x=window_size,
window_size_y=window_size,
direc_name=direc_name,
file_name_save=file_name_save,
training_direcs=training_direcs,
channel_names=channel_names,
num_of_features=2,
edge_feature=edge_feature,
dilation_radius=0,
border_mode="valid",
sample_mode="all",
output_mode="conv",
reshape_size=512,
display=False,
verbose=True,
process_std=False)
from deepcell import make_training_data_2d as make_training_data
# Define maximum number of training examples
max_training_examples = 1e6
window_size = 30
# Load data
direc_name = '/data/training_data/HeLa_joint/'
file_name_save = os.path.join('/data/training_data_npz/HeLa/', 'HeLa_joint_sample_61x61.npz')
training_direcs = ['set1', 'set2', 'set3', 'set4', 'set5']
channel_names = ['phase', 'nuclear']
# Specify the number of feature masks that are present
num_of_features = 2
# Specify which feature is the edge feature
edge_feature = [1,0,0]
# Create the training data
make_training_data(max_training_examples = max_training_examples, window_size_x = window_size, window_size_y = window_size,
direc_name = direc_name,
file_name_save = file_name_save,
training_direcs = training_direcs,
channel_names = channel_names,
num_of_features = 2,
edge_feature = edge_feature,
dilation_radius = 1,
display = False,
verbose = True)
channel_names = ["phase"] # Specify the number of feature masks that are present num_of_features = 2 # Specify which feature is the edge feature edge_feature = [1,0,0] # Create the training data make_training_data(max_training_examples = max_training_examples, window_size_x = window_size, window_size_y = window_size, direc_name = direc_name, file_name_save = file_name_save, training_direcs = training_direcs, channel_names = channel_names, num_of_features = 2, edge_feature = edge_feature, dilation_radius = 1, #sub_sample = True, sample_mode = "subsample", reshape_size = None, display = False, verbose = True, process_std = True)
from deepcell import make_training_data_movie as make_training_data
# Define maximum number of training examples
window_size = 30
# Load data
direc_name = '/data/DL_Training_Data/nuclear_movie'
file_name_save = os.path.join('/data/training_data_npz/nuclear_movie/', 'nuclear_movie_disc_same.npz')
training_direcs = ["set1", "set2"]
channel_names = ["DAPI"]
# Create the training data
make_training_data(window_size_x = 30, window_size_y = 30,
direc_name = direc_name,
file_name_save = file_name_save,
training_direcs = training_direcs,
channel_names = channel_names,
annotation_name = "corrected",
raw_image_direc = "RawImages",
annotation_direc = "Annotation",
border_mode = "same",
output_mode = "disc",
num_frames = 60,
reshaped_size = 256,
display = False,
num_of_frames_to_display = 5,
verbose = True)
"set1", "set2", "set3", "set4", "set5", "set6", "set7", "set8", "set9"
]
channel_names = ["nuclear"]
# Create output ditrectory, if necessary
pathlib.Path(output_directory).mkdir(parents=True, exist_ok=True)
# Specify the number of feature masks that are present
num_of_features = 2
# Specify which feature is the edge feature
edge_feature = [1, 0, 0]
# Create the training data
make_training_data(max_training_examples=max_training_examples,
window_size_x=window_size,
window_size_y=window_size,
direc_name=direc_name,
file_name_save=file_name_save,
training_direcs=training_direcs,
channel_names=channel_names,
num_of_features=2,
edge_feature=edge_feature,
dilation_radius=1,
border_mode="same",
output_mode="conv",
reshape_size=512,
display=False,
max_plotted=5,
verbose=True)
