cuda-pcl has some libraries used to process points cloud with CUDA and some samples for their usage. The are several subfolders in the project and every subfolder has:
- lib for segmentation implemented by CUDA
- Sample code can show the lib usage and also be used to check perf and accuracy by comparing its output with PCL
To get started, follow the instructions below. If you run into any issues please let us know.
To get started, follow these steps.
- Install Jetpack4.4.1 by SDKManager
- install PCL (Eigen included)
$sudo apt-get update
$sudo apt-get install libpcl-dev
Enter any subfolder and then
$ make
- Please boost CPU and GPU firstly
$ sudo nvpmodel -m 0
$ sudo jetson_clocks
- Usage:
$ ./demo [*.pcd]
$ strings lib* | grep version | grep lib
lib* version: 1.0 Jun 2 2019 09:30:19
Jetson Xavier AGX 8GB
Jetpack 4.4.1
CUDA 10.2
PCL 1.8
Eigen 3
https://developer.nvidia.com/blog/accelerating-lidar-for-robotics-with-cuda-based-pcl/