def _main():
"""Main script."""
args = default_config_parse(
default_config_path='settings/dataloader/KITTI.yaml')
config = Config.build_from_yaml(args.config)
configure_logging_verbosity(verbose=args.verbose)
save_labels_as_dataframe(**config.config)
def _main():
"""Main script."""
args = default_config_parse(
default_config_path='settings/scripts/compute_embedded_space.yaml')
configure_logging_verbosity(verbose=args.verbose)
config = Config.build_from_yaml(args.config)
save_embedded_features(**config.config)
def compute_feature_maps(output_path,
input_path,
inference_list_path,
resnet_config_path,
verbose=False):
"""Compute feature maps and save to disk.
The feature maps will be stored to output_path.
Args:
output_path (str or pathlib.Path): Path to save feature maps to.
input_path (str or pathlib.Path): Path to image folder.
inference_list_path (str or pathlib.Path): Path to inference list
(example creation and details at scripts/computer_inference_list.py).
resnet_config_path (str or pathlib.Path): Path to ResNet config file.
verbose (bool): Set verbosity.
"""
LOGGER.info("Compute feature maps ... ")
configure_logging_verbosity(verbose=verbose)
label_names = _get_label_names(inference_list_path)
image_paths = [input_path/(label_name + '.png') for label_name in label_names]
model = resnet.ResNet.build_from_yaml(resnet_config_path)
save_features(file_list=image_paths,
output_path=output_path,
model=model)
def _main():
"""Main script."""
args = default_config_parse(
default_config_path='settings/scripts/compute_inference_list.yaml')
configure_logging_verbosity(verbose=args.verbose)
config = Config.build_from_yaml(args.config)
compute_inference_list(**config.config, verbose=args.verbose)
def _main():
"""Main script."""
args = default_config_parse(
default_config_path=
'settings/scripts/plot_class_distribution_in_dataset.yaml')
configure_logging_verbosity(verbose=args.verbose)
config = Config.build_from_yaml(args.config)
plot_class_distribution(**config.config)
def compute_inference_list(label_path, output_path, seed=42, verbose=False):
"""Compute inference list and save to disk.
This method will save a .txt file with each column holding an unique identifier for a label.
For each class n amount of samples are written to the file. n is equal to the minimum amount of
samples for a class. For KITTI, Pedestrian_sitting is the class with the fewest occurrences
(222), so for every class 222 samples would be chosen.
Args:
label_path (str or pathlib.Path): Path to labels as pickled pandas Data Frame file.
output_path (str or pathlib.Path): Path to save the inference list to.
seed (int): Random seed to enable reproducibility.
verbose (True): Set verbosity.
"""
LOGGER.info("Compute inference list ... ")
configure_logging_verbosity(verbose=verbose)
random_state = np.random.RandomState(seed)
labels = pd.read_pickle(str(label_path))
n_samples_dict = dict()
# Count samples per class
for class_types in CLASS_LIST:
n_samples_dict[class_types] = np.sum(labels[TYPE] == class_types)
# From each class get the same amount of samples like the class with the fewest n of samples
min_n = n_samples_dict[min(n_samples_dict, key=n_samples_dict.get)]
inference_list = []
for class_types in CLASS_LIST:
labels_one_class = labels[labels[TYPE] == class_types]
identifier = random_state.choice(labels_one_class.index.values,
size=min_n,
replace=False)
inference_list.append(identifier)
inference_list = [item for sublist in inference_list for item in sublist]
np.savetxt(str(output_path), inference_list, fmt='%s')
def _main():
"""Main script."""
args = default_config_parse(default_config_path='settings/run.yaml')
configure_logging_verbosity(verbose=args.verbose)
config = Config.build_from_yaml(args.config)
_run_toolchain(config=config.config)