def load_dataset(self):
train = get_file_list(self.args.data_path, self.args.train_ids_path)
#val = get_file_list(self.args.data_path,self.args.valid_ids_path)
train, val = get_split_deterministic(train,
fold=0,
num_splits=5,
random_state=12345)
shapes = [brats_dataloader.load_patient(i)[0].shape[1:] for i in train]
max_shape = np.max(shapes, 0)
max_shape = list(np.max((max_shape, self.args.patch_size), 0))
dataloader_train = brats_dataloader(train,
self.args.batch_size,
max_shape,
self.args.num_threads,
return_incomplete=True,
infinite=False)
tr_transforms = get_train_transform(self.args.patch_size)
tr_gen = MultiThreadedAugmenter(dataloader_train,
tr_transforms,
num_processes=self.args.num_threads,
num_cached_per_queue=3,
seeds=None,
pin_memory=False)
self.num_batches_per_epoch = int(
math.ceil(len(train) / self.args.batch_size))
self.train_data_loader = tr_gen
self.val = val
tr_transforms.append(
GaussianBlurTransform(blur_sigma=(0.5, 1.5),
different_sigma_per_channel=True,
p_per_channel=0.5,
p_per_sample=0.15))
# now we compose these transforms together
tr_transforms = Compose(tr_transforms)
return tr_transforms
if __name__ == "__main__":
patients = get_list_of_patients(brats_preprocessed_folder)
train, val = get_split_deterministic(patients,
fold=0,
num_splits=5,
random_state=12345)
patch_size = (128, 128, 128)
batch_size = 2
# I recommend you don't use 'iteration oder all training data' as epoch because in patch based training this is
# really not super well defined. If you leave all arguments as default then each batch sill contain randomly
# selected patients. Since we don't care about epochs here we can set num_threads_in_multithreaded to anything.
dataloader = BraTS2017DataLoader3D(train, batch_size, patch_size, 1)
batch = next(dataloader)
try:
from batchviewer import view_batch
# batch viewer can show up to 4d tensors. We can show only one sample, but that should be sufficient here
view_batch(batch['data'][0], batch['seg'][0])
# 'patch_size': [24, 128, 128],
# 'pretrained': True
# }
# }
patients = get_list_of_patients(
'brats_data_preprocessed/Brats{}TrainingData'.format(
str(args.brats_train_year)))
batch_size = args.batch_size
patch_size = [args.patch_depth, args.patch_width, args.patch_height]
in_channels = ['t1c', 't2', 'flair']
#%%
# num_splits=5 means 1/5th is validation data!
patients_train, patients_val = get_split_deterministic(patients,
fold=0,
num_splits=5,
random_state=args.seed)
if not args.use_validation:
patients_train = patients
#%%
patients_test = get_list_of_patients(
'brats_data_preprocessed/Brats{}ValidationData'.format(
str(args.brats_test_year)))
target_patients = patients_test
#%%
train_dl = BRATSDataLoader(patients_train,
batch_size=batch_size,
patch_size=patch_size,
output_preprocessed_v0 = join(preprocessing_output_dir, task_name_variant0)
maybe_mkdir_p(output_preprocessed_v0)
patients = subdirs(join(root, "MR"), join=False)
task_name_variant1 = "Task037_CHAOS_Task_3_5_Variant1"
task_name_variant2 = "Task038_CHAOS_Task_3_5_Variant2"
output_preprocessed_v1 = join(preprocessing_output_dir, task_name_variant1)
maybe_mkdir_p(output_preprocessed_v1)
output_preprocessed_v2 = join(preprocessing_output_dir, task_name_variant2)
maybe_mkdir_p(output_preprocessed_v2)
splits = []
for fold in range(5):
tr, val = get_split_deterministic(patients, fold, 5, 12345)
train = ["T2_" + i
for i in tr] + ["T1_in_" + i
for i in tr] + ["T1_out_" + i for i in tr
] + ["CT_" + i for i in tr]
validation = ["T2_" + i for i in val] + ["T1_in_" + i for i in val] + [
"T1_out_" + i for i in val
] + ["CT_" + i for i in val]
splits.append({'train': train, 'val': validation})
save_pickle(splits, join(output_preprocessed_v0, "splits_final.pkl"))
splits = []
for fold in range(5):
tr, val = get_split_deterministic(patients, fold, 5, 12345)
train = ["T2_" + i for i in tr] + ["T1_" + i for i in tr]
validation = ["T2_" + i for i in val] + ["T1_" + i for i in val]
