def visu_caller(path, step_ind, relu_idx):
##########################
# Initiate the environment
##########################
# Choose which gpu to use
GPU_ID = '0'
# Set GPU visible device
os.environ['CUDA_VISIBLE_DEVICES'] = GPU_ID
# Disable warnings
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
###########################
# 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.
# No augmentation to avoid random inputs
config.augment_scale_anisotropic = False
config.augment_symmetries = [False, False, False]
config.augment_rotation = 'none'
config.augment_scale_min = 1.0
config.augment_scale_max = 1.0
config.augment_noise = 0.0
config.augment_occlusion = 'none'
config.augment_color = 1.0
config.batch_num = 2
config.in_radius = 5
##############
# Prepare Data
##############
print()
print('Dataset Preparation')
print('*******************')
# Initiate dataset configuration
if config.dataset.startswith('ModelNet40'):
dataset = ModelNet40Dataset(config.input_threads)
elif config.dataset == 'S3DIS':
dataset = S3DISDataset(config.input_threads)
on_val = True
elif config.dataset == 'Scannet':
dataset = ScannetDataset(config.input_threads, load_test=True)
elif config.dataset.startswith('ShapeNetPart'):
dataset = ShapeNetPartDataset(
config.dataset.split('_')[1], config.input_threads)
elif config.dataset == 'NPM3D':
dataset = NPM3DDataset(config.input_threads, load_test=True)
elif config.dataset == 'Semantic3D':
dataset = Semantic3DDataset(config.input_threads)
else:
raise ValueError('Unsupported dataset : ' + config.dataset)
# Create subsample clouds of the models
dl0 = config.first_subsampling_dl
dataset.load_subsampled_clouds(dl0)
# Initiate ERF input pipeleine (only diff is that it is not random)
dataset.init_ERF_input_pipeline(config)
##############
# Define Model
##############
print('Creating Model')
print('**************\n')
t1 = time.time()
if config.dataset.startswith('ShapeNetPart'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('S3DIS'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('Scannet'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('NPM3D'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('ModelNet40'):
model = KernelPointCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('Semantic3D'):
model = KernelPointFCNN(dataset.flat_inputs, config)
else:
raise ValueError('Unsupported dataset : ' + config.dataset)
# 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
visualizer = ModelVisualizer(model, restore_snap=chosen_snap)
t2 = time.time()
print('\n----------------')
print('Done in {:.1f} s'.format(t2 - t1))
print('----------------\n')
#####################
# Start visualization
#####################
print('Start visualization')
print('*******************\n')
visualizer.show_effective_recep_field(model, dataset, relu_idx)
else:
this_dataset = config.dataset
if plot_dataset:
if plot_dataset == this_dataset:
continue
else:
raise ValueError(
'All logs must share the same dataset to be compared')
else:
plot_dataset = this_dataset
# Plot the training loss and accuracy
compare_trainings(logs, logs_names)
# Plot the validation
if config.dataset_task == 'classification':
compare_convergences_classif(logs, logs_names)
elif config.dataset_task == 'cloud_segmentation':
if config.dataset.startswith('S3DIS'):
dataset = S3DISDataset(config, load_data=False)
compare_convergences_segment(dataset, logs, logs_names)
if config.dataset == 'APRPointCloud':
dataset = APRPointCloudDataset(config)
compare_convergences_segment(dataset, logs, logs_names)
elif config.dataset_task == 'slam_segmentation':
if config.dataset.startswith('SemanticKitti'):
dataset = SemanticKittiDataset(config)
compare_convergences_SLAM(dataset, logs, logs_names)
else:
raise ValueError('Unsupported dataset : ' + plot_dataset)
def visu_caller(path, step_ind, relu_idx, compute_activations):
# Check if activation have already been computed
if relu_idx is not None:
visu_path = os.path.join('visu', 'visu_' + path.split('/')[-1],
'top_activations',
'Relu{:02d}'.format(relu_idx))
if not os.path.exists(visu_path):
message = 'No activations found for Relu number {:d} of the model {:s}.'
print(message.format(relu_idx, path.split('/')[-1]))
compute_activations = True
else:
# Get the list of files
feature_files = np.sort(
[f for f in os.listdir(visu_path) if f.endswith('.ply')])
if len(feature_files) == 0:
message = 'No activations found for Relu number {:d} of the model {:s}.'
print(message.format(relu_idx, path.split('/')[-1]))
compute_activations = True
else:
compute_activations = True
if compute_activations:
##########################
# Initiate the environment
##########################
# Choose which gpu to use
GPU_ID = '0'
# Set GPU visible device
os.environ['CUDA_VISIBLE_DEVICES'] = GPU_ID
# Disable warnings
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '3'
###########################
# 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_symmetries = False
config.batch_num = 3
config.in_radius = 4
config.validation_size = 200
##############
# Prepare Data
##############
print()
print('Dataset Preparation')
print('*******************')
# Initiate dataset configuration
if config.dataset.startswith('ModelNet40'):
dataset = ModelNet40Dataset(config.input_threads)
elif config.dataset == 'S3DIS':
dataset = S3DISDataset(config.input_threads)
on_val = True
elif config.dataset == 'Scannet':
dataset = ScannetDataset(config.input_threads, load_test=True)
elif config.dataset.startswith('ShapeNetPart'):
dataset = ShapeNetPartDataset(
config.dataset.split('_')[1], config.input_threads)
elif config.dataset == 'NPM3D':
dataset = NPM3DDataset(config.input_threads, load_test=True)
elif config.dataset == 'Semantic3D':
dataset = Semantic3DDataset(config.input_threads)
else:
raise ValueError('Unsupported dataset : ' + config.dataset)
# Create subsample clouds of the models
dl0 = config.first_subsampling_dl
dataset.load_subsampled_clouds(dl0)
# Initialize input pipelines
if config.dataset == 'S3DIS':
dataset.init_input_pipeline(config)
else:
dataset.init_test_input_pipeline(config)
##############
# Define Model
##############
print('Creating Model')
print('**************\n')
t1 = time.time()
if config.dataset.startswith('ShapeNetPart'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('S3DIS'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('Scannet'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('NPM3D'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('ModelNet40'):
model = KernelPointCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('Semantic3D'):
model = KernelPointFCNN(dataset.flat_inputs, config)
else:
raise ValueError('Unsupported dataset : ' + config.dataset)
# 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
visualizer = ModelVisualizer(model, restore_snap=chosen_snap)
t2 = time.time()
print('\n----------------')
print('Done in {:.1f} s'.format(t2 - t1))
print('----------------\n')
#####################
# Start visualization
#####################
print('Start visualization')
print('*******************\n')
relu_idx = visualizer.top_relu_activations(model, dataset, relu_idx)
# Show the computed activations
ModelVisualizer.show_activation(path, relu_idx)
def test_caller(path, step_ind, on_val):
##########################
# Initiate the environment
##########################
# Choose which gpu to use
GPU_ID = '0'
# Set GPU visible device
os.environ['CUDA_VISIBLE_DEVICES'] = GPU_ID
# 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
if config.dataset.startswith('ModelNet40'):
dataset = ModelNet40Dataset(config.input_threads)
elif config.dataset == 'S3DIS':
dataset = S3DISDataset(config.input_threads)
on_val = True
elif config.dataset == 'Scannet':
dataset = ScannetDataset(config.input_threads, load_test=(not on_val))
elif config.dataset.startswith('ShapeNetPart'):
dataset = ShapeNetPartDataset(
config.dataset.split('_')[1], config.input_threads)
elif config.dataset == 'NPM3D':
dataset = NPM3DDataset(config.input_threads, load_test=(not on_val))
elif config.dataset == 'Semantic3D':
dataset = Semantic3DDataset(config.input_threads)
else:
raise ValueError('Unsupported dataset : ' + config.dataset)
# Create subsample clouds of the models
dl0 = config.first_subsampling_dl
dataset.load_subsampled_clouds(dl0)
# Initialize input pipelines
if on_val:
dataset.init_input_pipeline(config)
else:
dataset.init_test_input_pipeline(config)
##############
# Define Model
##############
print('Creating Model')
print('**************\n')
t1 = time.time()
if config.dataset.startswith('ShapeNetPart'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('S3DIS'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('Scannet'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('NPM3D'):
model = KernelPointFCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('ModelNet40'):
model = KernelPointCNN(dataset.flat_inputs, config)
elif config.dataset.startswith('Semantic3D'):
model = KernelPointFCNN(dataset.flat_inputs, config)
else:
raise ValueError('Unsupported dataset : ' + config.dataset)
# 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')
if config.dataset.startswith('ShapeNetPart'):
if config.dataset.split('_')[1] == 'multi':
tester.test_multi_segmentation(model, dataset)
else:
tester.test_segmentation(model, dataset)
elif config.dataset.startswith('S3DIS'):
tester.test_cloud_segmentation_on_val(model, dataset)
elif config.dataset.startswith('Scannet'):
if on_val:
tester.test_cloud_segmentation_on_val(model, dataset)
else:
tester.test_cloud_segmentation(model, dataset)
elif config.dataset.startswith('Semantic3D'):
if on_val:
tester.test_cloud_segmentation_on_val(model, dataset)
else:
tester.test_cloud_segmentation(model, dataset)
elif config.dataset.startswith('NPM3D'):
if on_val:
tester.test_cloud_segmentation_on_val(model, dataset)
else:
tester.test_cloud_segmentation(model, dataset)
elif config.dataset.startswith('ModelNet40'):
tester.test_classification(model, dataset)
else:
raise ValueError('Unsupported dataset')
###########################
# Load the model parameters
###########################
config = S3DISConfig()
##############
# Prepare Data
##############
print()
print('Dataset Preparation')
print('*******************')
# Initiate dataset configuration
dataset = S3DISDataset(config.input_threads)
# 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
##############
if plot_dataset:
if plot_dataset in config.dataset:
continue
else:
raise ValueError('All logs must share the same dataset to be compared')
else:
plot_dataset = config.dataset[:5]
# Plot the training loss and accuracy
compare_trainings(logs, logs_names)
# Plot the validation
if plot_dataset.startswith('Shape'):
compare_convergences_multisegment(logs, logs_names)
elif plot_dataset.startswith('S3DIS'):
dataset = S3DISDataset()
compare_convergences_segment(dataset, logs, logs_names)
elif plot_dataset.startswith('Model'):
dataset = ModelNet40Dataset()
compare_convergences_classif(dataset, logs, logs_names)
elif plot_dataset.startswith('Scann'):
dataset = ScannetDataset()
compare_convergences_segment(dataset, logs, logs_names)
elif plot_dataset.startswith('Seman'):
dataset = Semantic3DDataset()
compare_convergences_segment(dataset, logs, logs_names)
elif plot_dataset.startswith('NPM3D'):
dataset = NPM3DDataset()
compare_convergences_segment(dataset, logs, logs_names)
else:
raise ValueError('Unsupported dataset : ' + plot_dataset)
