def read_config(self):
config_path = self.config_path
cfg_from_yaml_file(self.config_path, cfg)
self.logger = common_utils.create_logger()
self.demo_dataset = DemoDataset(
dataset_cfg=cfg.DATA_CONFIG,
class_names=cfg.CLASS_NAMES,
training=False,
root_path=Path(
"/home/muzi2045/Documents/project/OpenPCDet/data/kitti/velodyne/000001.bin"
),
ext='.bin')
self.device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
self.net = build_network(model_cfg=cfg.MODEL,
num_class=len(cfg.CLASS_NAMES),
dataset=self.demo_dataset)
self.net.load_params_from_file(filename=self.model_path,
logger=self.logger,
to_cpu=True)
self.net = self.net.to(self.device).eval()
# nuscenes dataset
lidar2imu_t = np.array([0.985793, 0.0, 1.84019])
lidar2imu_r = Quaternion(
[0.706749235, -0.01530099378, 0.0173974518, -0.7070846])
self.lidar2imu = transform_matrix(lidar2imu_t,
lidar2imu_r,
inverse=True)
self.imu2lidar = transform_matrix(lidar2imu_t,
lidar2imu_r,
inverse=False)
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file', type=str, default='/home/syang/Data/data_object_velodyne/output/kitti_models/centernet_multihead/0124_single_head/centernet_multihead.yaml', help='specify the config for training')
parser.add_argument('--batch_size', type=int, default=1, required=False, help='batch size for training')
parser.add_argument('--workers', type=int, default=16, help='number of workers for dataloader')
parser.add_argument('--extra_tag', type=str, default='0129_twostage_first_4', help='extra tag for this experiment')
# os.path.abspath('..') + '/output/robosense_models/robosense_pointpillar/BResampl_LR001/ckpt/checkpoint_epoch_30.pth'
parser.add_argument('--ckpt', type=str, default='/home/syang/Data/data_object_velodyne/output/kitti_models/centernet_multihead/0124_single_head/ckpt/checkpoint_epoch_80.pth', help='checkpoint to start from')
parser.add_argument('--launcher', choices=['none', 'pytorch', 'slurm'], default='none')
parser.add_argument('--tcp_port', type=int, default=18888, help='tcp port for distrbuted training')
parser.add_argument('--local_rank', type=int, default=0, help='local rank for distributed training')
parser.add_argument('--set', dest='set_cfgs', default=None, nargs=argparse.REMAINDER,
help='set extra config keys if needed')
parser.add_argument('--max_waiting_mins', type=int, default=30, help='max waiting minutes')
parser.add_argument('--start_epoch', type=int, default=0, help='')
parser.add_argument('--eval_tag', type=str, default='default', help='eval tag for this experiment')
parser.add_argument('--eval_all', action='store_true', default=False, help='whether to evaluate all checkpoints')
parser.add_argument('--ckpt_dir', type=str, default=None, help='specify a ckpt directory to be evaluated if needed')
parser.add_argument('--save_to_file', action='store_true', default=False, help='')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
np.random.seed(1024)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument(
'--cfg_file',
type=str,
default='cfgs/kitti_models/second.yaml',
help='specify the config for demo') # 1. '--cfg_file' #指定配置
parser.add_argument('--data_path',
type=str,
default='demo_data',
help='specify the point cloud data file or directory'
) # 2 . '--data_path' #指定点云数据文件或目录
parser.add_argument(
'--ckpt', type=str, default=None,
help='specify the pretrained model') # 3. '--ckpt' #指定预训练模型
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file'
) # 4. '--ext' #指定点云数据文件的扩展名
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file,
cfg) # cfg的参数在tools/cfg/kitti_models/pv-rcnn.yaml
return args, cfg # cfg的参数在tools/cfg/kitti_models/pv-rcnn.yaml
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
#设置模型参数
parser.add_argument('--cfg_file',
type=str,
default='cfgs/kitti_models/second.yaml',
help='specify the config for demo')
#设置数据路径参数
parser.add_argument('--data_path',
type=str,
default='demo_data',
help='specify the point cloud data file or directory')
#设置预训练模型，Check point 校验点 在系统运行中当出现查找数据请求时，
#系统从数据库中找出这些数据并存入内存区，这样用户就可以对这些内存区数据进行修改等
parser.add_argument('--ckpt',
type=str,
default=None,
help='specify the pretrained model')
# extension 指定点云数据文件的扩展名
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
#arguments参数，configuration配制
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default=None,
help='specify the config for training')
parser.add_argument(
'--data_path',
type=str,
default='demo_data',
help='specify the dataset or point cloud data directory')
parser.add_argument('--ckpt',
type=str,
default=None,
help='checkpoint to start from')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
parser.add_argument('--save_video_path',
type=str,
default=None,
help='path to save the inference video')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(
args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
np.random.seed(1024)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default='cfgs/nuscenes_models/cbgs_pp_multihead.yaml',
help='specify the config for demo')
parser.add_argument('--data_path',
type=str,
default='/media/javier/HDD_linux/data/nuscenes',
help='specify the point cloud data file or directory')
parser.add_argument('--ckpt',
type=str,
default=None,
help='specify the pretrained model')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
parser.add_argument('--frames',
type=int,
default='6019',
help='specify the number of frames to use')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default='cfgs/kitti_models/second.yaml',
help='specify the config for demo')
parser.add_argument('--data_path',
type=str,
default='demo_data',
help='specify the point cloud data file or directory')
parser.add_argument('--ckpt',
type=str,
default=None,
help='specify the pretrained model')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file', type=str, default=None, help='specify the config for training')
parser.add_argument('--batch_size', type=int, default=None, required=False, help='batch size for training')
parser.add_argument('--workers', type=int, default=4, help='number of workers for dataloader')
parser.add_argument('--extra_tag', type=str, default='default', help='extra tag for this experiment')
parser.add_argument('--ckpt', type=str, default=None, help='checkpoint to start from')
parser.add_argument('--launcher', choices=['none', 'pytorch', 'slurm'], default='none')
parser.add_argument('--tcp_port', type=int, default=18888, help='tcp port for distrbuted training')
parser.add_argument('--local_rank', type=int, default=50, help='local rank for distributed training')
parser.add_argument('--set', dest='set_cfgs', default=None, nargs=argparse.REMAINDER,
help='set extra config keys if needed')
parser.add_argument('--max_waiting_mins', type=int, default=30, help='max waiting minutes')
parser.add_argument('--start_epoch', type=int, default=0, help='')
parser.add_argument('--eval_tag', type=str, default='default', help='eval tag for this experiment')
parser.add_argument('--eval_all', action='store_true', default=False, help='whether to evaluate all checkpoints')
parser.add_argument('--ckpt_dir', type=str, default=None, help='specify a ckpt directory to be evaluated if needed')
parser.add_argument('--save_to_file', action='store_true', default=False, help='')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
np.random.seed(1024)
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file', type=str, default=None, help='specify the config for training')
parser.add_argument('--batch_size', type=int, default=None, required=False, help='batch size for training')
parser.add_argument('--epochs', type=int, default=None, required=False, help='number of epochs to train for')
parser.add_argument('--workers', type=int, default=8, help='number of workers for dataloader')
parser.add_argument('--extra_tag', type=str, default='default', help='extra tag for this experiment')
parser.add_argument('--ckpt', type=str, default=None, help='checkpoint to start from')
parser.add_argument('--pretrained_model', type=str, default=None, help='pretrained_model')
parser.add_argument('--launcher', choices=['none', 'pytorch', 'slurm'], default='none')
parser.add_argument('--tcp_port', type=int, default=18888, help='tcp port for distrbuted training')
parser.add_argument('--sync_bn', action='store_true', default=False, help='whether to use sync bn')
parser.add_argument('--fix_random_seed', action='store_true', default=False, help='')
parser.add_argument('--ckpt_save_interval', type=int, default=1, help='number of training epochs')
parser.add_argument('--local_rank', type=int, default=0, help='local rank for distributed training')
parser.add_argument('--max_ckpt_save_num', type=int, default=30, help='max number of saved checkpoint')
parser.add_argument('--merge_all_iters_to_one_epoch', action='store_true', default=False, help='')
parser.add_argument('--set', dest='set_cfgs', default=None, nargs=argparse.REMAINDER,
help='set extra config keys if needed')
parser.add_argument('--max_waiting_mins', type=int, default=0, help='max waiting minutes')
parser.add_argument('--start_epoch', type=int, default=0, help='')
parser.add_argument('--save_to_file', action='store_true', default=False, help='')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs, cfg)
return args, cfg
def __init__(self, cfg_file, logger=False):
self.logger = False
if logger:
self.logger = True
logging.basicConfig(
filename="../results/log2.txt",
level=logging.INFO,
format='%(levelname)s: %(asctime)s %(message)s',
datefmt='%m/%d/%Y %I:%M:%S')
cfg_from_yaml_file(cfg_file, cfg)
self.model_dir = Path('../results')
self.dataset = JrdbDataset(
dataset_cfg=cfg.DATA_CONFIG,
class_names=['Pedestrian'],
root_path=None,
training=False,
logger=None,
)
self.num_frames = len(self.dataset.data_infos)
print(self.num_frames)
self.class_names = self.dataset.class_names
self.easy_eval = True
def setup_model(self):
cfg_from_yaml_file(self.detector_config, cfg)
self.logger = common_utils.create_logger()
self.dataset = DummyDataset(dataset_cfg=cfg.DATA_CONFIG,
class_names=cfg.CLASS_NAMES)
self.device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
self.net = build_network(model_cfg=cfg.MODEL,
num_class=len(cfg.CLASS_NAMES),
dataset=self.dataset)
self.net.load_params_from_file(filename=self.model_path,
logger=self.logger,
to_cpu=True)
self.net = self.net.to(self.device).eval()
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--bag_file', type=str, default=None, help='specify the bag file to be inferenced')
parser.add_argument('--cfg_file', type=str, default=None, help='specify the config for inference')
parser.add_argument('--save_video', default=False, action='store_true')
parser.add_argument('--save_path', default='../data/plusai/inference_result/', help='path to save the inference result')
parser.add_argument('--ckpt', type=str, default=None, help='model checkpoint')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
np.random.seed(1024)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument(
'--cfg_file',
type=str,
default=
'/home/syang/Data/data_object_velodyne/output/kitti_models/onenet_twostage_0130/test/onenet_twostage_0130.yaml',
help='specify the config for training')
parser.add_argument(
'--ckpt',
type=str,
default=
'/home/syang/Data/data_object_velodyne/output/kitti_models/onenet_twostage_0130/test/ckpt/checkpoint_epoch_78.pth',
help='checkpoint to start from')
parser.add_argument('--show_heatmap',
action='store_true',
default=False,
help='')
parser.add_argument('--batch_size',
type=int,
default=1,
required=False,
help='batch size for training')
parser.add_argument('--workers',
type=int,
default=4,
help='number of workers for dataloader')
parser.add_argument('--data_path',
type=str,
default='demo_data',
help='specify the point cloud data file or directory')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(
args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
np.random.seed(1024)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file', type=str, default=None, help='specify the config for training')
parser.add_argument('--ckpt', type=str, default=None, help='checkpoint to start from')
parser.add_argument('--data_path', type=str, default='demo_data',
help='specify the scene directory or val info pkl')
parser.add_argument('--save_path', default='../data/plusai/inference_result/', help='path to save the inference result')
parser.add_argument('--batch_size', type=int, default=1, required=False, help='batch size for training')
parser.add_argument('--workers', type=int, default=4, help='number of workers for dataloader')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
np.random.seed(1024)
return args, cfg
def __init__(self, input_dict):
# pvrcnn cfg
cfg_from_yaml_file(input_dict.cfg_file, cfg)
# create logger
log_dir = Path(str(input_dict.output_dir)) / 'log'
log_dir.mkdir(parents=True, exist_ok=True)
self.logger = common_utils.create_logger(
log_dir /
('log_%s.txt' % datetime.datetime.now().strftime('%Y%m%d-%H%M%S')),
rank=cfg.LOCAL_RANK)
# build dataset and network
self.demo_dataset = DemoDataset( # dummy dataset for preprocess inputdata
dataset_cfg=cfg.DATA_CONFIG,
class_names=cfg.CLASS_NAMES,
training=False,
root_path=input_dict.dummy_cloud,
ext='.bin',
logger=self.logger)
self.device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
self.model = build_network(model_cfg=cfg.MODEL,
num_class=len(cfg.CLASS_NAMES),
dataset=self.demo_dataset)
self.model.load_params_from_file(filename=input_dict.ckpt_file,
logger=self.logger,
to_cpu=self.device == "cpu")
self.model.to(self.device)
self.model.eval()
self.score_threshold = input_dict.score_threashold
# for ROS
self.action_server = actionlib.SimpleActionServer(
"excavator/lidar_perception/ros_pvrcnn_action",
detector3dAction,
execute_cb=self.execute_cb,
auto_start=False)
self.action_server.start()
self.mk_pub = rospy.Publisher("ros_pvrcnn", MarkerArray, queue_size=1)
self.cls_list = [String(cls) for cls in cfg.CLASS_NAMES]
def read_config(self):
config_path = self.config_path
cfg_from_yaml_file(self.config_path, cfg)
self.logger = common_utils.create_logger()
self.demo_dataset = DemoDataset(dataset_cfg=cfg.DATA_CONFIG,
class_names=cfg.CLASS_NAMES,
training=False,
root_path=Path("/none"),
ext='.bin')
self.device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
self.net = build_network(model_cfg=cfg.MODEL,
num_class=len(cfg.CLASS_NAMES),
dataset=self.demo_dataset)
print("Model path: ", self.model_path)
self.net.load_params_from_file(filename=self.model_path,
logger=self.logger,
to_cpu=True)
self.net = self.net.to(self.device).eval()
def read_config(self):
config_path = self.config_path
cfg_from_yaml_file(self.config_path, cfg)
self.logger = common_utils.create_logger()
self.demo_dataset = DemoDataset(
dataset_cfg=cfg.DATA_CONFIG,
class_names=cfg.CLASS_NAMES,
training=False,
root_path=Path(
"/home/muzi2045/Documents/project/OpenPCDet/data/kitti/velodyne/000001.bin"
),
ext='.bin')
self.device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
self.net = build_network(model_cfg=cfg.MODEL,
num_class=len(cfg.CLASS_NAMES),
dataset=self.demo_dataset)
self.net.load_params_from_file(filename=self.model_path,
logger=self.logger,
to_cpu=True)
self.net = self.net.to(self.device).eval()
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default='cfgs/kitti_models/second.yaml',
help='specify the config for demo')
# parser.add_argument('--data_path', type=str, default='demo_data',
# help='specify the point cloud data file or directory')
parser.add_argument(
'--data_root',
type=str,
default='demo_data',
help='specify the root of calib, velodyne and image files')
parser.add_argument('--file_number',
type=str,
default='000008',
help='specify file number to detect objects for')
parser.add_argument('--ckpt',
type=str,
default=None,
help='specify the pretrained model')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
# parser.add_argument('--ext')
parser.add_argument(
'--res',
type=str,
default="../results/3dod/vis/",
help="specify the results folder of the detection result")
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default='cfgs/kitti_models/second.yaml',
help='specify the config for demo')
parser.add_argument('--seq_path',
type=str,
default='demo_data',
help='specify the point cloud data sequence path')
parser.add_argument('--ckpt',
type=str,
default=None,
help='specify the pretrained model')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
parser.add_argument(
'--output_dir',
type=str,
default='',
help='The path to save predictions and output log files for tracking')
parser.add_argument(
'--saved_pred',
type=str,
default='',
help=
'The path to existing saved predictions and output log files for visualizing'
)
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default='cfgs/kitti_models/second.yaml',
help='specify the config for demo')
parser.add_argument('--data_path',
type=str,
default='demo_data',
help='specify the point cloud data file or directory')
parser.add_argument('--ckpt',
type=str,
default=None,
help='specify the pretrained model')
parser.add_argument(
'--ext',
type=str,
default='.bin',
help='specify the extension of your point cloud data file')
parser.add_argument('--vis',
type=bool,
default=False,
help='visualize detection results')
parser.add_argument('--point',
type=bool,
default=False,
help='save point prediction results')
parser.add_argument('--save_gt',
type=bool,
default=False,
help='save point ground truth labels')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
return args, cfg
def read_config(self):
print(self.config_path)
config_path = self.config_path
cfg_from_yaml_file(self.config_path, cfg)
self.logger = common_utils.create_logger()
self.demo_datasets = DemoDataset(
dataset_cfg=cfg.DATA_CONFIG,
class_names=cfg.CLASS_NAMES,
training=False,
root_path=Path(
'/home/syang/Data/RS_datasets/datasets/ruby119_longzhudadao_1200423181920/npy/ruby119_longzhudadao_1200423181920_755.npy'
),
ext='.npy')
self.device = torch.device(
"cuda" if torch.cuda.is_available() else "cpu")
print(self.device)
self.net = build_network(model_cfg=cfg.MODEL,
num_class=len(cfg.CLASS_NAMES),
dataset=self.demo_datasets)
self.net.load_params_from_file(filename=self.model_path,
logger=self.logger,
to_cpu=True)
self.net = self.net.to(self.device).eval()
model_path = 'cfgs/kitti_models/pointpillar_7728.pth'
config_path = 'cfgs/kitti_models/pp_multihead.yaml'
model_path = 'cfgs/kitti_models/pp_multihead_nds5823.pth'
'''
config_path = 'cfgs/kitti_models/second.yaml'
model_path = 'cfgs/kitti_models/second_7862.pth'
movelidarcenter = 0 #69.12/2
threshold = 0.4
proc_1 = Processor_ROS(config_path, model_path)
proc_1.initialize()
cfg_from_yaml_file(config_path, cfg)
'''
rospy.init_node('centerpoint_ros_node')
sub_lidar_topic = [ "/velodyne_points",
"/carla/ego_vehicle/lidar/lidar1/point_cloud",
"/kitti_player/hdl64e",
"/lidar_protector/merged_cloud",
"/merged_cloud",
"/lidar_top",
"/roi_pclouds",
"/livox/lidar",
"/SimOneSM_PointCloud_0"]
bag_frame_list = os.listdir(os.path.join(data_path, cur_bag_name, 'pointcloud'))
bag_frame_list.sort()
f.write(os.path.join(bag_name, 'pointcloud', bag_frame_list[int(pointcloud_idx[:-4]) + 1]) + '\n')
f.close()
if __name__ == '__main__':
parser = argparse.ArgumentParser()
parser.add_argument('--data_path', help='directory to data path which should contains bag and label')
parser.add_argument('--lidar_topic', default='/unified/lidar_points')
parser.add_argument('--odom_topic', default='/navsat/odom')
parser.add_argument('--cfg_file', type=str, default='/home/jingsen/workspace/OpenPCDet/tools/cfgs/livox_models/pv_rcnn_multiframe.yaml')
parser.add_argument('--visualize', action='store_true', default=False, help='visualize the multi-frame point cloud')
parser.add_argument('--num_workers', default=6, help='num workers to process label data')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
log_file = os.path.join(args.data_path, 'data_preprocessing_log.txt')
logger = create_logger(log_file, rank=0)
# logger.info('=== Start extract point-cloud and annotations from origin bag and label files, this will take a long time ... ===')
# preprocess_dataset()
logger.info('\n\n=== Start process multiframe dataset ... ===')
prepare_multiframe_dataset()
# logger.info('\n\n=== Start get image sets ... ===')
# get_images_sets()
print('log file saved in {}'.format(log_file))
def parse_config():
parser = argparse.ArgumentParser(description='KITTI Demo Video')
parser.add_argument('--maxdisp',
type=int,
default=192,
help='maxium disparity')
parser.add_argument('--loss_weights',
type=float,
nargs='+',
default=[0.25, 0.5, 1., 1.])
parser.add_argument('--max_disparity', type=int, default=192)
parser.add_argument('--maxdisplist',
type=int,
nargs='+',
default=[12, 3, 3])
parser.add_argument('--datatype', default='2015', help='datapath')
parser.add_argument('--datapath',
default='data/kitti/training',
help='datapath')
parser.add_argument('--epochs',
type=int,
default=300,
help='number of epochs to train')
parser.add_argument('--train_bsize',
type=int,
default=6,
help='batch size for training (default: 6)')
parser.add_argument('--test_bsize',
type=int,
default=8,
help='batch size for testing (default: 8)')
parser.add_argument('--save_path',
type=str,
default='results/pseudoLidar/',
help='the path of saving checkpoints and log')
parser.add_argument('--resume', type=str, default=None, help='resume path')
parser.add_argument('--lr', type=float, default=5e-4, help='learning rate')
parser.add_argument('--with_spn',
action='store_true',
help='with spn network or not')
parser.add_argument('--print_freq',
type=int,
default=5,
help='print frequence')
parser.add_argument('--init_channels',
type=int,
default=1,
help='initial channels for 2d feature extractor')
parser.add_argument('--nblocks',
type=int,
default=2,
help='number of layers in each stage')
parser.add_argument(
'--channels_3d',
type=int,
default=4,
help='number of initial channels 3d feature extractor ')
parser.add_argument('--layers_3d',
type=int,
default=4,
help='number of initial layers in 3d network')
parser.add_argument('--growth_rate',
type=int,
nargs='+',
default=[4, 1, 1],
help='growth rate in the 3d network')
parser.add_argument('--spn_init_channels',
type=int,
default=8,
help='initial channels for spnet')
parser.add_argument('--start_epoch_for_spn', type=int, default=121)
parser.add_argument(
'--pretrained',
type=str,
default='configs/checkpoint/kitti2015_ck/checkpoint.tar',
help='pretrained model path')
parser.add_argument('--split_file', type=str, default=None)
parser.add_argument('--evaluate', action='store_true')
parser.add_argument('--max_high', type=int, default=1)
parser.add_argument('--cfg_file',
type=str,
default=paper.cfg,
help='specify the config for demo')
parser.add_argument('--data_path', type=str, default='data/kitti/training')
parser.add_argument('--ckpt',
type=str,
default=paper.model,
help='specify the pretrained model')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
return args, cfg
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default=None,
help='specify the config for training')
parser.add_argument('--batch_size',
type=int,
default=None,
required=False,
help='batch size for training')
parser.add_argument('--epochs',
type=int,
default=None,
required=False,
help='number of epochs to train for')
parser.add_argument('--workers',
type=int,
default=8,
help='number of workers for dataloader')
parser.add_argument('--extra_tag',
type=str,
default='default',
help='extra tag for this experiment')
parser.add_argument('--ckpt',
type=str,
default=None,
help='checkpoint to start from')
parser.add_argument('--pretrained_model',
type=str,
default=None,
help='pretrained_model')
parser.add_argument('--launcher',
choices=['none', 'pytorch', 'slurm'],
default='none')
parser.add_argument('--tcp_port',
type=int,
default=18888,
help='tcp port for distrbuted training')
parser.add_argument('--sync_bn',
action='store_true',
default=False,
help='whether to use sync bn')
parser.add_argument('--fix_random_seed',
action='store_true',
default=False,
help='')
parser.add_argument('--ckpt_save_interval',
type=int,
default=1,
help='number of training epochs')
parser.add_argument('--local_rank',
type=int,
default=0,
help='local rank for distributed training')
parser.add_argument('--max_ckpt_save_num',
type=int,
default=30,
help='max number of saved checkpoint')
parser.add_argument('--merge_all_iters_to_one_epoch',
action='store_true',
default=False,
help='')
parser.add_argument('--set',
dest='set_cfgs',
default=None,
nargs=argparse.REMAINDER,
help='set extra config keys if needed')
parser.add_argument('--max_waiting_mins',
type=int,
default=0,
help='max waiting minutes')
parser.add_argument('--start_epoch', type=int, default=0, help='')
parser.add_argument('--save_to_file',
action='store_true',
default=False,
help='')
parser.add_argument('--adv',
action='store_true',
default=False,
help='adv defense or not')
parser.add_argument('--norm', type=str, default='inf', help='norm type')
parser.add_argument('--epsilon',
type=float,
default=0.01,
help='epsilon value')
parser.add_argument('--rec_type',
type=str,
default='both',
help='both: attack to points and reflectance'
'points: attack to points only'
'reflectance: attack to reflectance only')
parser.add_argument('--iterations',
type=int,
default=1,
help='iterations of different method')
parser.add_argument(
'--pgd',
type=bool,
default=False,
help=
'pgd adversarial type, when pgd is True, momentum should be False and iterations should be 10'
)
parser.add_argument(
'--momentum',
type=bool,
default=False,
help=
'adversarial type momentum, when momentum is True, pgd should be False and iterations should be 10'
)
parser.add_argument('--cfg_root_dir',
type=str,
default='',
help='model and relative informations save dir')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
cfg.EXP_GROUP_PATH = '/'.join(
args.cfg_file.split('/')[1:-1]) # remove 'cfgs' and 'xxxx.yaml'
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs, cfg)
return args, cfg
# with open(test_filename, 'wb') as f:
# pickle.dump(kitti_infos_test, f)
# print('Kitti info test file is saved to %s' % test_filename)
print('---------------Start create groundtruth database for data augmentation---------------')
if not cfg.DATA_CONFIG.TS_DATA:
dataset.set_split(train_split)
dataset.create_groundtruth_database(train_filename, split=train_split)
print('---------------Data preparation Done---------------')
if __name__ == '__main__':
if sys.argv.__len__() > 1 and sys.argv[1] == 'create_kitti_infos':
cfg_file = sys.argv[2]
cfg_from_yaml_file(cfg_file, cfg)
if not cfg.DATA_CONFIG.TS_DATA:
data_path = cfg.ROOT_DIR / 'data' / 'kitti'
save_path = cfg.ROOT_DIR / 'data' / 'kitti'
else:
data_path = cfg.ROOT_DIR / 'ts_data'
save_path = cfg.ROOT_DIR / 'ts_data'
create_kitti_infos(
data_path=data_path,
save_path=save_path
)
else:
A = KittiDataset(root_path='data/kitti', class_names=cfg.CLASS_NAMES, split='train', training=True)
import pdb
pdb.set_trace()
from pcdet.datasets.JRDB.jrdb_dataset import JrdbDataset
from pcdet.datasets.kitti.kitti_dataset import KittiDataset
from pcdet.config import cfg, cfg_from_list, cfg_from_yaml_file, log_config_to_file
from pathlib import Path
import pdb
import numpy as np
# root_path = Path('data/jrdb')
cfg_from_yaml_file('cfgs/kitti_models/pointrcnn_test.yaml', cfg)
dataset = JrdbDataset(
dataset_cfg=cfg.DATA_CONFIG,
class_names=['Pedestrian'],
root_path=None,
training=True,
logger=None,
)
dataset.set_split('train')
train_filename = (Path(__file__).resolve() / '../../'
).resolve() / 'data' / 'jrdb_temp' / ('jrdb_infos_train.pkl')
dataset.create_groundtruth_database(train_filename, split='train')
# cfg_from_yaml_file('cfgs/kitti_models/pointrcnn.yaml', cfg)
# dataset = KittiDataset(
# dataset_cfg=cfg.DATA_CONFIG,
# These locations are where the model weights and other misc files are stored.
MODEL_CFG_PATH = '/code/submission/OpenPCDet/tools/cfgs/waymo_models/pv_rcnn.yaml'
MODEL_WEIGHTS = '/code/submission/lib/wod_latency_submission/WAYMO_MODEL_WEIGHTS.pth'
# The names of the lidars and input fields that users might want to use for
# detection.
LIDAR_NAMES = ['TOP', 'REAR', 'FRONT', 'SIDE_LEFT', 'SIDE_RIGHT']
LIDAR_FIELD = 'RANGE_IMAGE_FIRST_RETURN'
# The data fields requested from the evaluation script should be specified in
# this field in the module.
DATA_FIELDS = [lidar_name + '_' + LIDAR_FIELD for lidar_name in LIDAR_NAMES]
# Global variables that hold the models and configurations.
model_cfg = cfg_from_yaml_file(MODEL_CFG_PATH, cfg)
logger = common_utils.create_logger()
dataset_processor = DatasetTemplate(dataset_cfg=model_cfg.DATA_CONFIG,
class_names=model_cfg.CLASS_NAMES,
training=False,
root_path=None,
logger=logger)
model = None
def initialize_model():
"""Method that will be called by the evaluation script to load the model and weights.
"""
global model
model = build_network(model_cfg=model_cfg.MODEL,
num_class=len(model_cfg.CLASS_NAMES),
def parse_config():
parser = argparse.ArgumentParser(description='arg parser')
parser.add_argument('--cfg_file',
type=str,
default=None,
help='specify the config for training')
parser.add_argument('--data_dir', type=str, default=None)
parser.add_argument('--batch_size',
type=int,
default=16,
required=False,
help='batch size for training')
parser.add_argument('--epochs',
type=int,
default=80,
required=False,
help='number of epochs to train for')
parser.add_argument('--workers',
type=int,
default=4,
help='number of workers for dataloader')
parser.add_argument('--extra_tag',
type=str,
default='default',
help='extra tag for this experiment')
parser.add_argument('--ckpt',
type=str,
default=None,
help='checkpoint to start from')
parser.add_argument('--pretrained_model',
type=str,
default=None,
help='pretrained_model')
parser.add_argument('--launcher',
choices=['none', 'pytorch', 'slurm'],
default='none')
parser.add_argument('--federated',
choices=['none', 'sync', 'async'],
default='none') #NOTE: for federated learning
parser.add_argument('--tcp_port',
type=int,
default=18888,
help='tcp port for distributed training')
parser.add_argument('--sync_bn',
action='store_true',
default=False,
help='whether to use sync bn')
parser.add_argument('--fix_random_seed',
action='store_true',
default=False,
help='whether to use sync bn')
parser.add_argument('--ckpt_save_interval',
type=int,
default=2,
help='number of training epochs')
parser.add_argument('--local_rank',
type=int,
default=0,
help='local rank for distributed training')
parser.add_argument('--max_ckpt_save_num',
type=int,
default=30,
help='max number of saved checkpoint')
parser.add_argument('--set',
dest='set_cfgs',
default=None,
nargs=argparse.REMAINDER,
help='set extra config keys if needed')
args = parser.parse_args()
cfg_from_yaml_file(args.cfg_file, cfg)
cfg.TAG = Path(args.cfg_file).stem
if args.set_cfgs is not None:
cfg_from_list(args.set_cfgs, cfg)
return args, cfg