def test_vol_loader():
return NPYDataLoader(**init_args['volume_image_data_loader']['val'])
def train_vol_loader():
return NPYDataLoader(**init_args['volume_image_data_loader']['train'])
factor=np.sqrt(0.1),
cooldown=0,
patience=10,
min_lr=1e-6)
early_stopper = EarlyStopping(monitor='val_loss',
min_delta=0.001,
patience=100)
csv_logger = CSVLogger('output/{}_{}_ctd.csv'.format(
datetime.datetime.now().isoformat(), title))
train_datagen = VolumeImageDataGenerator(
**init_args['volume_image_data_generator']['train']['init'])
val_datagen = VolumeImageDataGenerator(
**init_args['volume_image_data_generator']['val']['init'])
train_vol_loader = NPYDataLoader(
**init_args['volume_image_data_loader']['train'])
val_vol_loader = NPYDataLoader(**init_args['volume_image_data_loader']['val'])
train_iter_args = init_args['volume_image_data_generator']['train'][
'flow_from_loader']
train_iter_args['volume_image_data_loader'] = train_vol_loader
val_iter_args = init_args['volume_image_data_generator']['val'][
'flow_from_loader']
val_iter_args['volume_image_data_loader'] = val_vol_loader
image_shape = train_datagen.image_shape
regularization_factor = 1
model = Resnet3DBuilder.build_resnet_18(image_shape, nb_classes,
regularization_factor)
model.load_weights(weights)
compile_args = init_args['model']['compile']
init_args = yaml.load(stream)
except yaml.YAMLError as exc:
print(exc)
parser = argparse.ArgumentParser(description='Predict given weights')
parser.add_argument('weights', help='Trained weights')
args = parser.parse_args()
if args.weights:
weights = args.weights
title = re.sub('^output/resnet50_[a-z]+_', '', weights.strip('.h5'))
print("Prediction by {}.".format(title))
test_datagen = VolumeImageDataGenerator(
**init_args['volume_image_data_generator']['test']['init'])
test_vol_loader = NPYDataLoader(
**init_args['volume_image_data_loader']['test'])
test_iter_args = init_args['volume_image_data_generator']['test']['flow_from_loader']
test_iter_args['volume_image_data_loader'] = test_vol_loader
model = load_model(weights)
model_pred_args = init_args['model']['predict_generator']
model_pred_args['generator'] = test_datagen.flow_from_loader(
**test_iter_args)
predictions = model.predict_generator(**model_pred_args)
print(predictions.flatten())
print(len(predictions))
df_subm = pd.DataFrame(columns=['id', 'cancer'])
from keras.models import (load_model)
from preprocessing.volume_image import (VolumeImageDataGenerator)
from preprocessing.image_loader import (NPYDataLoader)
import yaml
with open("config.yml", 'r') as stream:
try:
config_args = yaml.load(stream)
except yaml.YAMLError as exc:
print(exc)
test_datagen = VolumeImageDataGenerator(
**config_args['volume_image_data_generator']['test'])
test_vol_loader = NPYDataLoader(
**config_args['volume_image_data_loader']['test'])
iterator_args = config_args['volume_image_data_generator']['flow_from_loader']
test_iter_args = iterator_args.copy()
test_iter_args['volume_image_data_loader'] = test_vol_loader
model = load_model('output/resnet34_stage1.h5')
df_subm = pd.DataFrame(columns=['id', 'cancer'])
for idx, fn in enumerate(test_vol_loader.filenames):
x = test_vol_loader.load(fn)
x = test_datagen.standardize(x)
x = x[np.newaxis, ...]
print("Predicting {} (batch shape: {})".format(fn, x.shape))
proba = model.test(x, batch_size=1)
print(proba)