def eval_for_uats_mc(model_dir,
model_name,
batch_size=1,
out_dir=None,
lesion=False,
eval=True):
GT_dir = '/cache/suhita/skin/preprocessed/labelled/test/'
print('create start')
img_arr, GT_arr = create_test_arrays(GT_dir)
print('create end')
DIM = img_arr.shape
from dataset_specific.skin_2D.model.uats_entropy import weighted_model
wm = weighted_model()
model = wm.build_model(img_shape=(DIM[1], DIM[2], DIM[3]),
learning_rate=1e-07,
gpu_id=None,
nb_gpus=None,
trained_model=os.path.join(model_dir,
model_name + '.h5'))
model[0].load_weights(os.path.join(model_dir, model_name + '.h5'))
val_supervised_flag = np.ones((DIM[0], DIM[1], DIM[2], 1), dtype='int8')
prediction = model[0].predict([img_arr, GT_arr, val_supervised_flag],
batch_size=batch_size,
verbose=1)
# weights epochs LR gpu_id dist orient prediction LRDecay earlyStop
evaluateFiles_arr(
img_path='/cache/suhita/skin/preprocessed/labelled/test/',
prediction=prediction,
connected_component=True,
out_dir=out_dir,
eval=eval,
lesion=lesion)
def eval_for_supervised(model_dir,
img_path,
model_name,
eval=True,
out_dir=None,
connected_component=True):
if eval:
img_arr, GT_arr = create_test_arrays(img_path, eval=eval)
else:
img_arr = create_test_arrays(img_path, eval=eval)
DIM = img_arr.shape
from dataset_specific.skin_2D.model.baseline import weighted_model
wm = weighted_model()
model = wm.build_model(img_shape=(DIM[1], DIM[2], DIM[3]),
learning_rate=1e-07)
model.load_weights(os.path.join(model_dir, model_name + '.h5'))
prediction = model.predict(img_arr, batch_size=1, verbose=1)
# weights epochs LR gpu_id dist orient prediction LRDecay earlyStop
evaluateFiles_arr(img_path=img_path,
prediction=prediction,
connected_component=connected_component,
out_dir=out_dir,
eval=eval)
## Parse arguments
parser = argparse.ArgumentParser()
parser.add_argument('-g', '--gpu_num', type=str, default='2', help='GPU Number')
parser.add_argument('-f', '--fold_num', type=int, default=1, help='Fold Number')
parser.add_argument('-e', '--ens_folder_name', type=str, help='ensemble folder name')
parser.add_argument('-d', '--ds', type=str, default=SKIN_DATASET_NAME, help='dataset name')
config = tf.compat.v1.ConfigProto()
config.gpu_options.allow_growth = True
config.allow_soft_placement = True
try:
args = parser.parse_args()
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu_num
metadata = get_metadata(args.ds)
# Build Model
wm = weighted_model()
train(None, None,
dataset_name=args.ds,
ens_folder_name=args.temp_path,
labelled_perc=args.perc,
fold_num=args.fold_num,
model_type=wm
)
finally:
if os.path.exists(metadata[m_root_temp_path] + args.temp_path):
cleanup(metadata[m_root_temp_path] + args.temp_path)
print('clean up done!')