def do_registration_eval(args, registration_pair_list, pair_name_list=None):
"""
set running env and run the task
registration_pair_list includes source_path_list, target_path_list, l_source_path_list, l_target_path_list
source_path_list: the source image list, each item refers to the abstract path of the image
target_path_list: the target image list, each item refers to the abstract path of the image
l_source_path_list: optional, the label of source image list, each item refers to the abstract path of the image
l_target_path_list:optional, the label of target image list, each item refers to the abstract path of the image
:param args: the parsed arguments
:param registration_pair_list: list of registration pairs, [source_list, target_list, lsource_list, ltarget_list]
:return: None
"""
task_output_path = args.task_output_path
os.makedirs(task_output_path, exist_ok=True)
setting_folder_path = args.setting_folder_path
dm, tsm = init_test_env(setting_folder_path, task_output_path,
registration_pair_list, pair_name_list)
tsm.task_par['tsk_set']['gpu_ids'] = args.gpu_id
model_path = args.model_path
if model_path is not None:
assert os.path.isfile(model_path), "the model {} not exist".format_map(
model_path)
tsm.task_par['tsk_set'][
'model_path'] = model_path #if not tsm.task_par['tsk_set']['train']:
force_test_setting(dm, tsm, task_output_path)
dm_json_path = os.path.join(
task_output_path, 'cur_data_setting.json') if dm is not None else None
tsm_json_path = os.path.join(task_output_path, 'cur_task_setting.json')
run_one_task(tsm_json_path, dm_json_path)
def do_segmentation_eval(args, segmentation_file_list):
"""
set running env and run the task
:param args: the parsed arguments
:param segmentation_file_list: list of segmentation file list, [image_list, label_list]
:return: None
"""
task_output_path = args.task_output_path
os.makedirs(task_output_path, exist_ok=True)
setting_folder_path = args.setting_folder_path
file_txt_path = args.file_txt_path
fname_txt_path = file_txt_path.replace("file_path_list.txt",
"file_name_list.txt")
fname_list = read_txt_into_list(fname_txt_path) if os.path.isfile(
fname_txt_path) else None
dm, tsm = init_test_env(setting_folder_path, task_output_path,
segmentation_file_list, fname_list)
tsm.task_par['tsk_set']['gpu_ids'] = args.gpu_id
model_path = args.model_path
if model_path is not None:
assert os.path.isfile(model_path), "the model {} not exist".format_map(
model_path)
tsm.task_par['tsk_set']['model_path'] = model_path
force_test_setting(dm, tsm, task_output_path)
dm_json_path = os.path.join(
task_output_path, 'cur_data_setting.json') if dm is not None else None
tsm_json_path = os.path.join(task_output_path, 'cur_task_setting.json')
run_one_task(tsm_json_path, dm_json_path)
def __do_segmentation_train(args, pipeline=None):
"""
set running env and run the task
:param args: the parsed arguments
:param pipeline:a Pipeline object
:return: a Pipeline object
"""
output_root_path = args.output_root_path
task_name = args.task_name
data_task_name = args.data_task_name
setting_folder_path = args.setting_folder_path
data_task_path = os.path.join(output_root_path, data_task_name)
task_output_path = os.path.join(data_task_path, task_name)
os.makedirs(task_output_path, exist_ok=True)
dm, tsm = init_train_env(setting_folder_path, output_root_path, task_name,
data_task_name)
tsm.task_par['tsk_set']['gpu_ids'] = args.gpu_id
dm_json_path = os.path.join(
task_output_path, 'cur_data_setting.json') if dm is not None else None
tsm_json_path = os.path.join(task_output_path, 'cur_task_setting.json')
tsm.save(tsm_json_path)
if dm is not None:
dm.save(dm_json_path)
data_loaders = pipeline.data_loaders if pipeline is not None else None
pipeline = run_one_task(tsm_json_path, dm_json_path, data_loaders)
return pipeline
def __do_registration_train(args, pipeline=None):
"""
set running env and run the task
:param args: the parsed arguments
:param pipeline:a Pipeline object, only used for two-stage training, the pipeline of the first stage (including dataloader) would be pass to the second stage
:return: a Pipeline object
"""
output_root_path = args.output_root_path
task_name = args.task_name
data_task_name = args.data_task_name
setting_folder_path = args.setting_folder_path
data_task_path = os.path.join(output_root_path, data_task_name)
task_output_path = os.path.join(data_task_path, task_name)
os.makedirs(task_output_path, exist_ok=True)
dm, tsm = init_train_env(setting_folder_path, output_root_path, task_name,
data_task_name)
dm, tsm = addition_settings_for_two_stage_training(dm, tsm)
tsm.task_par['tsk_set']['gpu_ids'] = args.gpu_id
dm_json_path = os.path.join(
task_output_path, 'cur_data_setting.json') if dm is not None else None
tsm_json_path = os.path.join(task_output_path, 'cur_task_setting.json')
tsm.save(tsm_json_path)
if dm is not None:
dm.save(dm_json_path)
data_loaders = pipeline.data_loaders if pipeline is not None else None
pipeline = run_one_task(tsm_json_path, dm_json_path, data_loaders)
return pipeline
def do_registration_eval(args, registration_pair_list, pair_name_list=None):
"""
set running env and run the task
registration_pair_list includes source_path_list, target_path_list, l_source_path_list, l_target_path_list
source_path_list: the source image list, each item refers to the abstract path of the image
target_path_list: the target image list, each item refers to the abstract path of the image
l_source_path_list: optional, the label of source image list, each item refers to the abstract path of the image
l_target_path_list:optional, the label of target image list, each item refers to the abstract path of the image
:param args: the parsed arguments
:param registration_pair_list: list of registration pairs, [source_list, target_list, lsource_list, ltarget_list]
:return: None
"""
task_output_path = args.task_output_path
os.makedirs(task_output_path, exist_ok=True)
run_demo = args.run_demo
if run_demo:
demo_name = args.demo_name
if demo_name not in ['ants', 'demons', 'nifty_reg']:
setting_folder_path = os.path.join('./demo_settings/mermaid',
demo_name)
else:
setting_folder_path = os.path.join('./demo_settings', demo_name)
assert os.path.isdir(
setting_folder_path
), "the {} is not in supported demo list, eval_network_rdmm/eval_network_vsvf/opt_vsvf/opt_rdmm/opt_rdmm_predefined/ants/demons/niftyreg"
# task_output_path = os.path.join('./demo_output/mermaid',demo_name)
else:
setting_folder_path = args.setting_folder_path
dm, tsm = init_test_env(setting_folder_path, task_output_path,
registration_pair_list, pair_name_list)
tsm.task_par['tsk_set']['gpu_ids'] = args.gpu_id
# if not tsm.task_par['tsk_set']['train']:
force_test_setting(dm, tsm, task_output_path)
dm_json_path = os.path.join(
task_output_path, 'cur_data_setting.json') if dm is not None else None
tsm_json_path = os.path.join(task_output_path, 'cur_task_setting.json')
run_one_task(tsm_json_path, dm_json_path)
def __do_segmentation_train(self, pipeline=None):
"""
set running env and run the task
:param pipeline:a Pipeline object
:return: a Pipeline object
"""
self.pipeline = pipeline
self.output_root_path = self.args.output_root_path
self.task_name = self.args.task_name
self.data_task_name = self.args.data_task_name
self.setting_folder_path = self.args.setting_folder_path
self.data_task_path = os.path.join(output_root_path,data_task_name)
self.task_output_path = os.path.join(data_task_path,task_name)
os.makedirs(task_output_path, exist_ok=True)
dm, tsm = self.init_train_env()
tsm.task_par['tsk_set']['gpu_ids'] = args.gpu_id
self.dm_json_path = os.path.join(task_output_path, 'cur_data_setting.json') if dm is not None else None
self.tsm_json_path = os.path.join(task_output_path, 'cur_task_setting.json')
tsm.save(self.tsm_json_path)
if dm is not None:
dm.save(self.dm_json_path)
data_loaders = pipeline.data_loaders if self.pipeline is not None else None
self.pipeline = run_one_task(self.tsm_json_path, self.dm_json_path, data_loaders)