Wrong labels due to new Semantic Segmentation Classes

[KevinLADLee]

Due to changes after 0.9.14, 'vehicle' type not work, which may cause wrong labels.

[KevinLADLee]

carla_dataset_tools/recorder/world.py

Lines 18 to 42 in 453b9bc

	def save_to_disk(self, frame_id, timestamp, debug=False):
	# TODO: Save all object bbox in world
	# Frame Timestamp CityObjectLabel carla_id location rotation box_location box_extent
	object_labels = []
	object_labels += self.get_env_objects_labels(frame_id, timestamp, carla.CityObjectLabel.Vehicles)
	object_labels += self.get_env_objects_labels(frame_id, timestamp, carla.CityObjectLabel.Pedestrians)
	carla_actors = self.carla_world.get_actors()
	for carla_actor in carla_actors:
	if carla_actor.type_id.startswith('vehicle') \
	or carla_actor.type_id.startswith('walker'):
	transform = carla_transform_to_transform(carla_actor.get_transform())
	bbox = carla_bbox_to_bbox(carla_actor.bounding_box)
	if carla_actor.type_id.startswith('walker'):
	label_type = 'pedestrian'
	else:
	label_type = 'vehicle'
	object_labels.append(ObjectLabel(frame=frame_id,
	timestamp=timestamp,
	label_type=label_type,
	carla_id=carla_actor.id,
	transform=transform,
	bounding_box=bbox))