def main(args):
config = Config(args.in_settings)
config.init(args.dataset)
if args.verbose:
config.debug()
in_dir1 = config.temp_2d_dir
in_dir2 = config.temp_3d_dir
out_dir = config.temp_ensemble_dir
analyze_ensemble(in_dir1=in_dir1,
in_dir2=in_dir2,
out_dir=out_dir,
verbose=args.verbose)
def main(args):
config = Config(args.in_settings)
config.init(args.dataset)
if args.connectivity != 0:
config.ensemble_connectivity = args.connectivity
if args.verbose:
config.debug()
in_dir = config.temp_ensemble_dir
out_dir = config.output_path
connectivity = config.ensemble_connectivity
analyze_connection(in_dir=in_dir,
out_dir=out_dir,
in_filename=args.in_filename,
stat=args.stat,
connectivity=connectivity,
verbose=args.verbose)
def predict(in_settings, overwrite_dataset=None, verbose=False):
import tensorflow as tf
config = tf.ConfigProto(gpu_options=tf.GPUOptions(
allow_growth=True, per_process_gpu_memory_fraction=0.8))
session = tf.Session(config=config)
keras.backend.tensorflow_backend.set_session(session)
# config
config = Config(in_settings)
config.init(overwrite_dataset)
if verbose:
config.debug()
# loading sample
ds = Dataset2d()
width, height, imgnum = ds.load_csv(
os.path.join(config.input_path, 'input.csv'))
model_mn = 16 # model magick number
width_ex = int(math.ceil(width / model_mn) * model_mn)
height_ex = int(math.ceil(height / model_mn) * model_mn)
x_train = ds.image_read(os.path.join(config.input_path, '*.tif'), width,
height)
x_train_norm, x_width, x_height = reshape(x_train, (width_ex, height_ex))
# create model
model_base = Model2d(config)
model = model_base.load(input_shape=(width_ex, height_ex, 1))
if verbose:
print('x_train_norm:', x_train_norm.shape)
print(model.summary())
# predict
y_pred = model.predict(x_train_norm,
batch_size=config.predict_2d_batch_size,
verbose=1 if verbose else 0)
output = postprocess(y_pred, (width_ex, height_ex), (x_width, x_height))
# saving
os.makedirs(config.temp_2d_dir, exist_ok=True)
ds.image_save(output, config.temp_2d_dir)
def main(args):
config = Config(args.in_settings)
config.init(args.dataset)
if args.verbose:
config.debug()
sample_dir = config.input_path
model_path = config.model_3d_path
out_dir = config.temp_3d_dir
batch_size = config.predict_3d_batch_size
crop_size = config.predict_3d_crop_size
overlap = config.predict_3d_overlap
predict(sample_dir=sample_dir,
model_path=model_path,
out_dir=out_dir,
batch_size=batch_size,
crop_size=crop_size,
overlap=overlap,
application=args.application,
verbose=args.verbose)
