es = EarlyStopping(monitor="val_loss",
min_delta=1e-4,
patience=10,
verbose=1,
mode='auto')
callbacks_list = [checkpoint, tb, es]
######### PREPROCESS TRAINING DATA #########
DATA_DIR = os.path.join("data", "train")
PREPROCESSED_DIR = os.path.join(DATA_DIR, "preprocessed")
SKULLSTRIP_SCRIPT_PATH = os.path.join("utils", "CT_BET.sh")
preprocess.preprocess_dir(
DATA_DIR,
PREPROCESSED_DIR,
SKULLSTRIP_SCRIPT_PATH,
)
######### DATA IMPORT #########
ct_patches, mask_patches = patch_ops.CreatePatchesForTraining(
atlasdir=PREPROCESSED_DIR,
plane=plane,
patchsize=PATCH_SIZE,
max_patch=num_patches,
num_channels=num_channels)
print("Individual patch dimensions:", ct_patches[0].shape)
print("Num patches:", len(ct_patches))
print("ct_patches shape: {}\nmask_patches shape: {}".format(
ct_patches.shape, mask_patches.shape))
PREPROCESSED_DIR = os.path.join(VAL_DIR, "preprocess")
if not os.path.exists(PREPROCESSED_DIR):
os.makedirs(PREPROCESSED_DIR)
############### MODEL SELECTION ###############
with open(results.model) as json_data:
model = model_from_json(json.load(json_data))
model.load_weights(results.weights)
############### PREPROCESSING ###############
classes = results.classes.replace(" ", "").split(',')
preprocess_dir(VAL_DIR, PREPROCESSED_DIR, classes, results.numcores)
# get class encodings
class_encodings = get_classes(classes)
print(class_encodings)
############### DATA IMPORT ###############
patch_size = tuple([int(x) for x in results.patch_size.split('x')])
X, y, filenames, num_classes, img_shape = load_patch_data(
PREPROCESSED_DIR,
patch_size=patch_size,
num_patches=results.num_patches,
classes=classes)
############### PREDICT ###############