def set_up_config_dict(config_file, train_subjects, test_subjects,
job_identifier, batch_size_test, dataset_path,
input_to_get, output_to_get, train_subjects_model,
test_subjects_model, config_file_model,
job_identifier_encdec, epoch_encdec):
config_dict_for_saved_model = rhodin_utils_io.loadModule(
config_file_model).config_dict
config_dict_for_saved_model[
'implicit_rotation'] = config_dict_for_saved_model.get(
'implicit_rotation', False)
config_dict_for_saved_model[
'skip_background'] = config_dict_for_saved_model.get(
'skip_background', True)
config_dict_for_saved_model[
'loss_weight_pose3D'] = config_dict_for_saved_model.get(
'loss_weight_pose3D', 0)
config_dict_for_saved_model[
'n_hidden_to3Dpose'] = config_dict_for_saved_model.get(
'n_hidden_to3Dpose', 2)
config_dict_for_saved_model['train_subjects'] = train_subjects_model
config_dict_for_saved_model['test_subjects'] = test_subjects_model
config_dict_module = rhodin_utils_io.loadModule(config_file)
config_dict = config_dict_module.config_dict
config_dict_for_saved_model['job_identifier'] = job_identifier_encdec
config_dict['job_identifier_encdec'] = job_identifier_encdec
config_dict['job_identifier'] = job_identifier
config_dict['train_subjects'] = train_subjects
config_dict['test_subjects'] = test_subjects
config_dict['data_dir_path'] = dataset_path
config_dict['dataset_folder_train'] = dataset_path
config_dict['dataset_folder_test'] = dataset_path
config_dict['dataset_folder_mocap'] = os.path.join(
dataset_path, 'treadmill_lameness_mocap_ci_may11/mocap/')
config_dict['dataset_folder_rgb'] = os.path.join(dataset_path,
'animals_data/')
root = dataset_path.rsplit('/', 2)[0]
config_dict['bg_folder'] = config_dict_for_saved_model['bg_folder']
config_dict['rot_folder'] = config_dict_for_saved_model['rot_folder']
config_dict['pretrained_network_path'] = tep.get_model_path(
config_dict_for_saved_model, epoch=epoch_encdec)
assert os.path.exists(config_dict['pretrained_network_path'])
for val in input_to_get:
if val not in config_dict['input_types']:
config_dict['input_types'].append(val)
config_dict['output_types'] = []
for val in output_to_get:
if val not in config_dict['output_types']:
config_dict['output_types'].append(val)
return config_dict
def set_up_config_dict(config_path,
train_subjects,
test_subjects,
job_identifier,
batch_size_test,
dataset_path,
input_to_get,
output_to_get):
config_dict = rhodin_utils_io.loadModule(config_path).config_dict
config_dict['job_identifier'] = job_identifier
config_dict['train_subjects'] = train_subjects
config_dict['test_subjects'] = test_subjects
config_dict['implicit_rotation'] = config_dict.get('implicit_rotation', False)
config_dict['skip_background'] = config_dict.get('skip_background', True)
config_dict['loss_weight_pose3D'] = config_dict.get('loss_weight_pose3D', 0)
config_dict['n_hidden_to3Dpose'] = config_dict.get('n_hidden_to3Dpose', 2)
config_dict['batch_size_test'] = batch_size_test
config_dict['data_dir_path'] = dataset_path
config_dict['dataset_folder_train'] = dataset_path
config_dict['dataset_folder_test'] = dataset_path
# config_dict['bg_folder'] = '../data/median_bg/'
# config_dict['rot_folder'] = '../data/rotation_cal_1/'
for val in input_to_get:
if val not in config_dict['input_types']:
config_dict['input_types'].append(val)
for val in output_to_get:
if val not in config_dict['output_types']:
config_dict['output_types'].append(val)
return config_dict
def main(args):
train_subjects = re.split('/', args.train_subjects)
test_subjects = re.split('/', args.test_subjects)
train_subjects_model = re.split('/', args.train_subjects_model)
test_subjects_model = re.split('/', args.test_subjects_model)
config_dict_for_saved_model = rhodin_utils_io.loadModule(
args.config_file_model).config_dict
config_dict_for_saved_model[
'implicit_rotation'] = config_dict_for_saved_model.get(
'implicit_rotation', False)
config_dict_for_saved_model[
'skip_background'] = config_dict_for_saved_model.get(
'skip_background', True)
config_dict_for_saved_model[
'loss_weight_pose3D'] = config_dict_for_saved_model.get(
'loss_weight_pose3D', 0)
config_dict_for_saved_model[
'n_hidden_to3Dpose'] = config_dict_for_saved_model.get(
'n_hidden_to3Dpose', 2)
config_dict_for_saved_model['train_subjects'] = train_subjects_model
config_dict_for_saved_model['test_subjects'] = test_subjects_model
config_dict_module = rhodin_utils_io.loadModule(args.config_file)
config_dict = config_dict_module.config_dict
config_dict_for_saved_model['job_identifier'] = args.job_identifier_encdec
config_dict['job_identifier_encdec'] = args.job_identifier_encdec
config_dict['job_identifier'] = args.job_identifier
config_dict['train_subjects'] = train_subjects
config_dict['test_subjects'] = test_subjects
config_dict['data_dir_path'] = args.dataset_path
config_dict['dataset_folder_train'] = args.dataset_path
config_dict['dataset_folder_test'] = args.dataset_path
root = args.dataset_path.rsplit('/', 2)[0]
config_dict['pretrained_network_path'] = tedp.get_model_path(
config_dict_for_saved_model, epoch=args.epoch_encdec)
config_dict['rot_folder'] = config_dict_for_saved_model['rot_folder']
config_dict['bg_folder'] = config_dict_for_saved_model['bg_folder']
ignite = IgniteTestPainFromLatent(config_dict_module.__file__,
config_dict,
config_dict_for_saved_model,
epoch=args.epoch)
if ignite.model is not None:
ignite.run()
def set_up_config_dict(config_path, train_subjects, test_subjects,
job_identifier, job_identifier_encdec, batch_size_test,
dataset_path):
config_dict_module = rhodin_utils_io.loadModule(config_path)
config_dict = config_dict_module.config_dict
config_dict['job_identifier_encdec'] = job_identifier_encdec
config_dict['job_identifier'] = job_identifier
config_dict['train_subjects'] = train_subjects
config_dict['test_subjects'] = test_subjects
config_dict['batch_size_test'] = batch_size_test
config_dict['data_dir_path'] = dataset_path
config_dict['dataset_folder_train'] = dataset_path
config_dict['dataset_folder_test'] = dataset_path
config_dict['implicit_rotation'] = config_dict.get('implicit_rotation',
False)
config_dict['skip_background'] = config_dict.get('skip_background', True)
config_dict['loss_weight_pose3D'] = config_dict.get(
'loss_weight_pose3D', 0)
config_dict['n_hidden_to3Dpose'] = config_dict.get('n_hidden_to3Dpose', 2)
return config_dict
help="Job identifier for the saved model to load.")
parser.add_argument('--epoch_encdec',
type=str,
help="Which epoch for the saved model to load.")
return parser.parse_args(argv)
if __name__ == "__main__":
args = parse_arguments(sys.argv[1:])
print(args)
train_subjects = re.split('/', args.train_subjects)
test_subjects = re.split('/', args.test_subjects)
train_subjects_model = re.split('/', args.train_subjects_model)
test_subjects_model = re.split('/', args.test_subjects_model)
config_dict_for_saved_model = rhodin_utils_io.loadModule(
args.config_file_model).config_dict
config_dict_for_saved_model[
'implicit_rotation'] = config_dict_for_saved_model.get(
'implicit_rotation', False)
config_dict_for_saved_model[
'skip_background'] = config_dict_for_saved_model.get(
'skip_background', True)
config_dict_for_saved_model[
'loss_weight_pose3D'] = config_dict_for_saved_model.get(
'loss_weight_pose3D', 0)
config_dict_for_saved_model[
'n_hidden_to3Dpose'] = config_dict_for_saved_model.get(
'n_hidden_to3Dpose', 2)
config_dict_for_saved_model['train_subjects'] = train_subjects_model
config_dict_for_saved_model['test_subjects'] = test_subjects_model
type=str,
help="Slurm job ID, or other identifier, to not overwrite output.")
parser.add_argument('--resume',
type=int,
default=0,
help="epoch num to resume from")
return parser.parse_args(argv)
if __name__ == "__main__":
args = parse_arguments(sys.argv[1:])
# print(args)
train_subjects = re.split('/', args.train_subjects)
test_subjects = re.split('/', args.test_subjects)
# print (args.config_file)
config_dict_module = rhodin_utils_io.loadModule(args.config_file)
config_dict = config_dict_module.config_dict
config_dict['job_identifier'] = args.job_identifier
config_dict['train_subjects'] = train_subjects
config_dict['test_subjects'] = test_subjects
config_dict['data_dir_path'] = args.dataset_path
config_dict['dataset_folder_train'] = args.dataset_path
config_dict['dataset_folder_test'] = args.dataset_path
root = args.dataset_path.rsplit('/', 2)[0]
print(config_dict['test_subjects'])
ignite = IgniteTrainNVS(config_dict_module.__file__, config_dict,
args.resume)
# if ignite is not None:
ignite.run()
# config_dict_module.__file__, config_dict)
df_reduced.reset_index(drop=True)
return df_reduced
def get_label_df_for_subjects(data_folder, subjects):
subject_fi_dfs = []
print('Iterating over frame indices per subject (.csv files)')
for subject in subjects:
subject_frame_index_dataframe = pd.read_csv(data_folder + subject + '_reduced_frame_index.csv')
subject_fi_dfs.append(subject_frame_index_dataframe)
frame_index_df = pd.concat(subject_fi_dfs, ignore_index=True)
return frame_index_df
if __name__ == '__main__':
config_dict_module = rhodin_utils_io.loadModule("configs/config_pain_debug.py")
config_dict = config_dict_module.config_dict
print (config_dict['save_every'])
train_subjects = ['aslan','brava','herrera','julia','kastanjett','naughty_but_nice','sir_holger']
config_dict['train_subjects'] = train_subjects
config_dict['dataset_folder_train'] = '/local_storage/users/sbroome/SLU_LPS/pain_no_pain_x2h_intervals_for_extraction_128_128_2fps/'
dataset = SimpleFrameDataset(data_folder=config_dict['dataset_folder_train'],
subjects = config_dict['train_subjects'],
input_types=config_dict['input_types'],
label_types=config_dict['label_types_train'])
sampler = SimpleRandomFrameSampler(
data_folder=config_dict['dataset_folder_train'],
subjects=config_dict['train_subjects'],
views=config_dict['views'],
every_nth_frame=config_dict['every_nth_frame'])
# str_aft = '_'.join(['','reduced','thresh','%.2f'%float(thresh),'frame','index'])+'.csv'
# else:
# str_aft = '_reduced_frame_index.csv'
for subject in subjects:
csv_file = os.path.join(data_folder, subject + str_aft)
subject_frame_index_dataframe = pd.read_csv(csv_file)
# print (csv_file,len(subject_frame_index_dataframe))
subject_fi_dfs.append(subject_frame_index_dataframe)
frame_index_df = pd.concat(subject_fi_dfs, ignore_index=True)
return frame_index_df
if __name__ == '__main__':
config_dict_module = rhodin_utils_io.loadModule(
"configs/config_train_rotation_bl.py")
config_dict = config_dict_module.config_dict
print(config_dict['save_every'])
train_subjects = [
'aslan', 'brava', 'herrera', 'inkasso', 'julia', 'kastanjett',
'naughty_but_nice', 'sir_holger'
]
config_dict['train_subjects'] = train_subjects
dataset = MultiViewDataset(data_folder=config_dict['dataset_folder_train'],
bg_folder=config_dict['bg_folder'],
input_types=config_dict['input_types'],
label_types=config_dict['label_types_train'],
subjects=config_dict['train_subjects'],
rot_folder=config_dict['rot_folder'])
# batch_sampler = MultiViewDatasetSampler(
