def __init__(self,
nusc: NuScenes,
config: DetectionConfig,
result_path: str,
eval_set: str,
output_dir: str = None,
verbose: bool = True):
"""
Initialize a NuScenesEval object.
:param nusc: A NuScenes object.
:param config: A DetectionConfig object.
:param result_path: Path of the nuScenes JSON result file.
:param eval_set: The dataset split to evaluate on, e.g. train or val.
:param output_dir: Folder to save plots and results to.
:param verbose: Whether to print to stdout.
"""
self.nusc = nusc
self.result_path = result_path
self.eval_set = eval_set
self.output_dir = output_dir
self.verbose = verbose
self.cfg = config
# Make dirs
self.plot_dir = os.path.join(self.output_dir, 'plots')
if not os.path.isdir(self.output_dir):
os.makedirs(self.output_dir)
if not os.path.isdir(self.plot_dir):
os.makedirs(self.plot_dir)
# Load data
self.pred_boxes = load_prediction(self.result_path,
self.cfg.max_boxes_per_sample,
verbose=verbose)
self.gt_boxes = load_gt(self.nusc, self.eval_set, verbose=verbose)
assert set(self.pred_boxes.sample_tokens) == set(self.gt_boxes.sample_tokens), \
"Samples in split doesn't match samples in predictions."
# Add center distances
self.pred_boxes = add_center_dist(nusc, self.pred_boxes)
self.gt_boxes = add_center_dist(nusc, self.gt_boxes)
# Filter boxes (distance, points per box, etc.)
if verbose:
print('=> Filtering predictions')
self.pred_boxes = filter_eval_boxes(nusc,
self.pred_boxes,
self.cfg.class_range,
verbose=verbose)
if verbose:
print('=> Filtering ground truth annotations')
self.gt_boxes = filter_eval_boxes(nusc,
self.gt_boxes,
self.cfg.class_range,
verbose=verbose)
self.sample_tokens = self.gt_boxes.sample_tokens
def __init__(
self,
*,
data_root=const.DEFAULT_DATA_ROOT,
split="train",
version="v1.0",
full_dataset: bool = False,
download=False,
sensor_data: List[str] = SENSOR_DATA_KEYS,
coordinates="global",
):
self.root = os.path.join(data_root, NUSCENES_LOCAL_PATH)
if version not in VERSIONS:
raise ValueError(
f"version provided was {version} but only valid versions are: "
f"{VERSIONS}")
if full_dataset is False:
split = f"mini_{split}"
version = f"{version}-mini"
else:
version = f"{version}-trainval"
if download:
self.download(version=version)
nu = NuScenes(dataroot=self.root, version=version)
self.nu = nu
self.scenes = nu.scene
self.split = split
self.sample_tokens = loaders.load_gt(nusc=nu,
eval_split=split).sample_tokens
for s in sensor_data:
if s not in SENSOR_DATA_KEYS:
raise ValueError(f"sensor key: {s} is not a valid sensor. "
f"Valid sensors are: {SENSOR_DATA_KEYS}")
self.data_keys = sensor_data
if coordinates not in COORDINATES:
raise ValueError(f"coordinates can only be one of {COORDINATES} "
f"but {self.coordinates} was given.")
self.coordinates = coordinates