Closed rfsch closed 5 years ago
Currently the official version doesn't support this.
There are two ways to realize it.
Use synchronous mode and check if a point is in a bounding box of vehicles(using the location of previous frame). Otherwise it is a ground point. There are some known issues related to this, like #1647, #1312 . So you may have to do some adaption.
If you are familiar with the source code, you can get the actor id when lidar sensor hit a object. The actor id is available via
auto actor_temp = HitInfo->GetActor();
auto actor_id = GetEpisode().FindActor(actor_temp).GetActorId();
In carla-0.9.5/Unreal/CarlaUE4/Plugins/Carla/Source/Carla/Sensor/RayCastLidar.cpp
Or you may check my forked branch at mul_t
Thank you for your answer. But how would it be possible to get in this way Lane and road border markings. I do not want to have to detect or discover them programmatically. I would need to have the information about them in xyz in order to reason about them.
Thanks
@rfsch Sorry, seems I don't quite get your question.
If you are seeking for a groundtruth road, check this document.
save_to_disk
provides the whole map in opendrive format.
get_waypoint
provides informations about roads. Example of using get_waypoint API could be found in the no_rendering_mode.py.
Hope it helps.
@user025 thank you for your help!
Sorry if I didn't explain it well.
I would need LIDAR recordings of a car (Carla?) driving along a road. For detection of cars that are oncoming to the car I am driving I am using Lane borders. For these I would need to correct xyz coordinates and knowledge that they are lane borders.
Could I record the lidar data captured and know which points are borsers and lanes ?
Thanks!
@rfsch Lidar data provides points only. You can get groundtruth of borders and lanes of lidar data either by bounding box checking, as I mentioned in my previous answers, or by manually labeling.
@user025 So does it mean that by getting the actor id of an object hit by the lidar sensor I am able know it is a lane, border, etc ? I would need this information while driving and would not want to manually label it. Or is manually labeling easier ?
@rfsch actor id will always return 0 if lidar hit at anything not an carla actor, etc. buildings, grass, road. I believe manually labeling would be easier.
@user025 thanks. That was what I was afraid of. How would I manually label in Carla ?
https://github.com/ZhuangYanDLUT/carla have you checked this ?
@barbierParis Thanks! This looks good.
@rfsch if you use this forked version, would you mind giving some feedback on the results ? I'm currently having issues running CARLA on linux so I haven't been able to try it.
This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.
Hello
I would like to use Carla with LIDAR sensor information. Is there annotated road marking data available in pointcloud format. I.e. is it possible to have together with a Lidar frame (xyz data) ground thruth of the road boarder, lanes together in xyz format ?
Thanks