Ask Your Question
0

Lidar cluster and classifier not output as expectation

asked 2021-09-22 04:31:49 -0600

daniel.tian gravatar image

After run lidar cluster and classifier module(Single lidar) of autoware.auto, I found that the object output is not my expectation. I expect the lidar algo can output object only for interest ones: car/truck/bus/bicycle/person etc. But in fact it will output all the cluster points, the points from kerb/bushes/wall will also generate some dummy object. Of course, these objects is not what I need. So I want to know whether autoware algo can output only interest object, if it is, what shall I do? Below is the gif for reference. image description

edit retag flag offensive close merge delete

1 Answer

Sort by ยป oldest newest most voted
1

answered 2021-09-22 08:34:56 -0600

Hi @daniel.tian

You need to process the point cloud produced by the lidar with a separate algorithm, as per documentation VoxelNet and LMNet are compatible:

"lidar_detector reads point cloud data from 3D laser scanners, and provides LiDAR-based object detection capabilities. The basic performance comes from the Euclidean Clustering algorithm, which finds clusters of the LiDAR scan (point cloud) above the ground. To classify the clusters, DNN-based algorithms are also supported, such as VoxelNet and LMNet." https://gitlab.com/autowarefoundation...

edit flag offensive delete link more

Comments

Thanks for your reply Omar. My project is based on autoware.auto framework, and searched in my project, it seems that there is no VoxelNet and LMNet in official autoware.auto project. I guess it is located in autoware.ai. If so, after copy from autoware.ai, can I build it and launch in autoware.auto directly? I am new to autoware, and not know the work load. Many thanks.

daniel.tian gravatar image daniel.tian  ( 2021-09-22 22:24:37 -0600 )edit
1

You will have to process the point cloud on a separate node. Here is an example using ROS https://github.com/AbangLZU/VoxelNetRos

Glad it helped. Pls remember to vote and if you have more questions start a new one. Good luck

osilva gravatar image osilva  ( 2021-09-23 03:05:52 -0600 )edit

Your Answer

Please start posting anonymously - your entry will be published after you log in or create a new account.

Add Answer

Question Tools

Stats

Asked: 2021-09-22 04:31:49 -0600

Seen: 31 times

Last updated: Sep 22 '21