def nuscenes_data_prep(root_path,
version,
dataset_name='NuScenesDataset',
max_sweeps=10,
use_flat_vehicle_coords=False,
use_second_format_direction=False):
assert not use_flat_vehicle_coords, 'use_flat_vehicle_coords is not supported anymore.'
assert not use_second_format_direction, 'use_second_format_direction is not supported anymore.'
# Note: replace 'splits' to 'train', 'val' or 'test'
postfix = 'zero_slope' if use_flat_vehicle_coords else 'slope'
filename_template = f'infos_splits_{postfix}_sweepsN.pkl'
filename_template = filename_template.replace('N', str(max_sweeps))
nu_ds.create_nuscenes_infos(root_path,
version=version,
filename_template=filename_template,
max_sweeps=max_sweeps,
use_flat_vehicle_coords=use_flat_vehicle_coords,
use_second_format_direction=use_second_format_direction)
# filename_template = f'infos_train_slope_sweeps{max_sweeps}_12.pkl'
split = 'test' if 'test' in version else 'train'
name = filename_template.replace('splits', split)
create_groundtruth_database(
dataset_class_name=dataset_name,
data_path=root_path,
info_path=Path(root_path) / name,
db_dirname=f'gt_database_sweeps{max_sweeps}',
)
def nuscenes_data_prep(root_path, version, dataset_name, max_sweeps=10):
nu_ds.create_nuscenes_infos(root_path,
version=version,
max_sweeps=max_sweeps)
name = "infos_train.pkl"
if version == "v1.0-test":
name = "infos_test.pkl"
create_groundtruth_database(dataset_name, root_path,
Path(root_path) / name)
def nuscenes_data_prep(root_path, version, dataset_name, max_sweeps=10):
root_path = '/mnt/sdd/jhyoo/dataset/NUSCENES'
home_path = Path('/home/spalab/jskim_2/second.pytorch/second/dataset')
# nu_ds.create_nuscenes_infos(root_path, version=version, max_sweeps=max_sweeps)
name = "infos_train.pkl"
if version == "v1.0-test":
name = "infos_test.pkl"
database_save_path = home_path / 'gt_database'
db_info_save_path = home_path / "kitti_dbinfos_train.pkl"
create_groundtruth_database(dataset_class_name=dataset_name,
data_path=root_path,
info_path=home_path / name,
database_save_path=database_save_path,
db_info_save_path=db_info_save_path)
def nuscenes_data_prep(root_path, version, max_sweeps=10):
nu_ds.create_nuscenes_infos(root_path,
version=version,
max_sweeps=max_sweeps)
name = "infos_train.pkl"
if version == "v1.0-test":
name = "infos_test.pkl"
if max_sweeps == 0:
create_groundtruth_database("NuScenesDataset", root_path,
Path(root_path) / "infos_train.pkl")
else:
print(
"WARNING: ground truth database will be disabled because sweeps don't support this."
)
def nuscenes_data_prep(root_path, version, dataset_name, max_sweeps=10):
root_path = '/mnt/sdd/jhyoo/dataset/NUSCENES'
home_path = Path('/home/spalab/jskim_2/second.pytorch/second/dataset')
modality = 'radar'
radar_version = 'version2'
nu_rad_ds.create_nuscenes_infos(root_path,
version=version,
modality=modality,
radar_version=radar_version,
max_sweeps=max_sweeps)
name = f"infos_train_multisweep_{radar_version}.pkl"
if version == "v1.0-test":
name = "infos_test.pkl"
database_save_path = home_path / f'gt_database_multisweep_{radar_version}'
db_info_save_path = home_path / f"kitti_dbinfos_multisweep_{radar_version}_train.pkl"
create_groundtruth_database(dataset_class_name=dataset_name,
data_path=root_path,
info_path=Path(root_path) / name,
database_save_path=database_save_path,
db_info_save_path=db_info_save_path)
def kitti_gt_fgm_data_prep(old_root_path, old_trainval_info_path,
new_root_path, new_train_info_path):
from second.core import box_np_ops
from second.data.dataset import get_dataset_class
dataset = get_dataset_class('KittiDataset')(
root_path=old_root_path, info_path=old_trainval_info_path)
for i in trange(len(dataset)):
image_idx = i
sensor_data = dataset.get_sensor_data(i)
if 'image_idx' in sensor_data['metadata']:
image_idx = sensor_data['metadata']['image_idx']
points = sensor_data['lidar']['points']
annos = sensor_data['lidar']['annotations']
gt_boxes = annos['boxes']
gt_mask = box_np_ops.points_in_rbbox(points, gt_boxes)
points_aug = np.concatenate(
(points, gt_mask.max(axis=1, keepdims=True)), axis=1)
points_aug = points_aug.astype(np.float32)
velo_file = 'training/velodyne_reduced/%06d.bin' % (image_idx)
with open(f'{new_root_path}/{velo_file}', 'w') as f:
points_aug.tofile(f)
create_groundtruth_database(dataset_class_name='KittiFGMDataset',
data_path=new_root_path,
info_path=new_train_info_path)
def kitti_data_prep(root_path):
kitti_ds.create_kitti_info_file(root_path)
kitti_ds.create_reduced_point_cloud(root_path)
create_groundtruth_database("KittiDataset", root_path,
Path(root_path) / "kitti_infos_train.pkl")
def nuscenes_data_prep(root_path, version):
nu_ds.create_nuscenes_infos(root_path, version=version)
create_groundtruth_database("NuScenesDataset", root_path, Path(root_path) / "infos_train.pkl")
def appolo_data_prep(root_path="/scratch2/wdong/appolo/data"):
appolo_ds.create_kitti_info_file(root_path)
create_groundtruth_database("AppoloDataset", root_path,
Path(root_path) / "kitti_infos_train.pkl")
pass
def udi_data_prep(root_path, dataset_name):
udi_ds.create_udi_infos(root_path)
name = "infos_udi_train.pkl"
create_groundtruth_database(dataset_name, root_path,
Path(root_path) / name)
def lyft_data_prep(root_path, json_path, dataset_name, max_sweeps=10):
lyft_ds.create_lyft_infos(root_path, json_path, max_sweeps=max_sweeps)
name = "infos_train.pkl"
create_groundtruth_database(dataset_name, root_path,
Path(root_path) / name)
