def load_data_test(self, config_dict):
dataset = collected_dataset.CollectedDataset(
data_folder=config_dict['dataset_folder_test'],
img_type=config_dict['img_type'],
input_types=config_dict['input_types'],
label_types=config_dict['label_types_test'])
batch_sampler = collected_dataset.CollectedDatasetSampler(
data_folder=config_dict['dataset_folder_test'],
useSubjectBatches=0,
useCamBatches=config_dict['useCamBatches'],
batch_size=config_dict['batch_size_test'],
randomize=True,
every_nth_frame=100) #config_dict['every_nth_frame'])
loader = torch.utils.data.DataLoader(
dataset,
batch_sampler=batch_sampler,
num_workers=config_dict['num_workers'],
pin_memory=False,
collate_fn=utils_data.default_collate_with_string)
if 0: # save data for demo
import pickle
data_iterator = iter(loader)
data_input, data_labels = next(
data_iterator) #[next(data_iterator) for i in range(3)]
batch_size = 8
input = {
'img':
np.array(data_input['img'][:batch_size].numpy(),
dtype='float16'),
'bg':
np.array(data_input['bg'][:batch_size].numpy(),
dtype='float16'),
'R_cam_2_world':
np.array(data_input['R_cam_2_world'][:batch_size].numpy(),
dtype='float16'),
}
label = {
'3D':
np.array(data_labels['3D'][:batch_size].numpy(),
dtype='float16'),
'pose_mean':
np.array(data_labels['pose_mean'][:batch_size].numpy(),
dtype='float16'),
'pose_std':
np.array(data_labels['pose_std'][:batch_size].numpy(),
dtype='float16')
}
data_cach = tuple([input, label])
pickle.dump(data_cach, open('../examples/test_set.pickl', "wb"))
IPython.embed()
exit()
return loader
def load_data_test(self,config_dict):
dataset = collected_dataset.CollectedDataset(data_folder=config_dict['dataset_folder_test'],
input_types=config_dict['input_types'], label_types=config_dict['label_types_test'])
batch_sampler = collected_dataset.CollectedDatasetSampler(data_folder=config_dict['dataset_folder_test'],
useSubjectBatches=0, useCamBatches=config_dict['useCamBatches'],
batch_size=config_dict['batch_size_test'],
randomize=True,
every_nth_frame=config_dict['every_nth_frame'])
loader = torch.utils.data.DataLoader(dataset, batch_sampler=batch_sampler, num_workers=0, pin_memory=False,
collate_fn=utils_data.default_collate_with_string)
return loader
def load_data_test(self, config_dict):
#factory = dataset_factory.DatasetFactory()
#testloader = factory.load_data_test(config_dict_test)
dataset = collected_dataset.CollectedDataset(
data_folder=
'/cvlabdata1/home/rhodin/code/humanposeannotation/python/pytorch_human_reconstruction/TMP/H36M-MultiView-test',
input_types=config_dict['input_types'],
label_types=config_dict['label_types_test'],
useSubjectBatches=0,
useCamBatches=config_dict['useCamBatches'],
randomize=False)
testloader = torch.utils.data.DataLoader(
dataset,
batch_size=config_dict['batch_size_test'],
shuffle=False,
num_workers=config_dict['num_workers'],
pin_memory=False,
drop_last=True,
collate_fn=utils_data.default_collate_with_string)
testloader = utils_data.PostFlattenInputSubbatchTensor(testloader)
return testloader
def load_data_train(self, config_dict):
#return load_data_test(config_dict) # HACK
#factory = dataset_factory.DatasetFactory()
#trainloader, valloader_UNUSED = factory.load_data_train(config_dict_cams)
dataset = collected_dataset.CollectedDataset(
data_folder=
'/cvlabdata1/home/rhodin/code/humanposeannotation/python/pytorch_human_reconstruction/TMP/H36M-MultiView-train',
input_types=config_dict['input_types'],
label_types=config_dict['label_types_train'],
useSubjectBatches=config_dict['useSubjectBatches'],
useCamBatches=config_dict['useCamBatches'], # HACK
useSequentialFrames=config_dict.get('useSequentialFrames', 0),
randomize=True)
trainloader = torch.utils.data.DataLoader(
dataset,
batch_size=config_dict['batch_size_train'],
shuffle=True,
num_workers=config_dict['num_workers'],
pin_memory=False,
drop_last=True,
collate_fn=utils_data.default_collate_with_string)
trainloader = utils_data.PostFlattenInputSubbatchTensor(trainloader)
return trainloader