###########################
# Load the model parameters
###########################
config = ThreeDMatchConfig()
##############
# Prepare Data
##############
print()
print('Dataset Preparation')
print('*******************')
# Initiate dataset configuration
dataset = ThreeDMatchDataset(config.input_threads,
voxel_size=config.first_subsampling_dl)
# Create subsampled input clouds
dl0 = config.first_subsampling_dl
# dataset.load_subsampled_clouds(dl0)
# Initialize input pipelines
dataset.init_input_pipeline(config)
# Test the input pipeline alone with this debug function
# dataset.check_input_pipeline_timing(config)
##############
# Define Model
##############
def test_caller(path, step_ind, on_val):
# Disable warnings
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '0'
###########################
# Load the model parameters
###########################
# Load model parameters
config = Config()
config.load(path)
##################################
# Change model parameters for test
##################################
# Change parameters for the test here. For example, you can stop augmenting the input data.
#config.augment_noise = 0.0001
#config.augment_color = 1.0
#config.validation_size = 500
#config.batch_num = 10
##############
# Prepare Data
##############
print()
print('Dataset Preparation')
print('*******************')
# Initiate dataset configuration
dataset = ThreeDMatchDataset(1, load_test=True)
# Initialize input pipelines
dataset.init_test_input_pipeline(config)
##############
# Define Model
##############
print('Creating Model')
print('**************\n')
t1 = time.time()
model = KernelPointFCNN(dataset.flat_inputs, config)
# Find all snapshot in the chosen training folder
snap_path = os.path.join(path, 'snapshots')
snap_steps = [
int(f[:-5].split('-')[-1]) for f in os.listdir(snap_path)
if f[-5:] == '.meta'
]
# Find which snapshot to restore
chosen_step = np.sort(snap_steps)[step_ind]
chosen_snap = os.path.join(path, 'snapshots',
'snap-{:d}'.format(chosen_step))
# Create a tester class
tester = ModelTester(model, restore_snap=chosen_snap)
t2 = time.time()
print('\n----------------')
print('Done in {:.1f} s'.format(t2 - t1))
print('----------------\n')
############
# Start test
############
print('Start Test')
print('**********\n')
tester.generate_descriptor(model, dataset)
config_path = f'snapshot/{chosen_snap}/config.json'
config = json.load(open(config_path, 'r'))
config = edict(config)
config.architecture = [
'simple',
'resnetb',
]
for i in range(config.num_layers - 1):
config.architecture.append('resnetb_strided')
config.architecture.append('resnetb')
config.architecture.append('resnetb')
for i in range(config.num_layers - 1):
config.architecture.append('nearest_upsample')
config.architecture.append('unary')
dset = ThreeDMatchDataset(root='/ssd2/xuyang/3DMatch',
downsample=0.05,
config=config)
dataloader, neighborhood_limits = get_dataloader(dset,
batch_size=1,
num_workers=10)
import pdb
pdb.set_trace()
import time
s_time = time.time()
for iter, inputs in enumerate(dataloader):
print(iter)
if iter == 1000:
break
e_time = time.time()
print(f"{e_time - s_time:.2f}s")
# momentum=config.momentum,
weight_decay=config.weight_decay,
)
config.scheduler = optim.lr_scheduler.ExponentialLR(
config.optimizer,
gamma=config.scheduler_gamma,
)
# create dataset and dataloader
train_set = ThreeDMatchDataset(root=config.root,
split='train',
downsample=config.downsample,
self_augment=config.self_augment,
num_node=config.num_node,
augment_noise=config.augment_noise,
augment_axis=config.augment_axis,
augment_rotation=config.augment_rotation,
augment_translation=config.augment_translation,
config=config,
)
val_set = ThreeDMatchDataset(root=config.root,
split='val',
num_node=64,
downsample=config.downsample,
self_augment=config.self_augment,
augment_noise=config.augment_noise,
augment_axis=config.augment_axis,
augment_rotation=config.augment_rotation,
augment_translation=config.augment_translation,
config=config,