Build and install Intel's librealsense for the NVIDIA Jetson Nano Developer Kit
MIT License
Build and install scripts for Intel's librealsense for the NVIDIA Jetson Nano Developer Kit
Original article on JetsonHacks: https://wp.me/p7ZgI9-34j
The Intel® RealSense™ SDK is here: https://github.com/IntelRealSense/librealsense The SDK library name is librealsense. This is for version 2 of the library, which supports the D400 series depth cameras, T265 tracking camera, and the SR300 depth camera.
It is now possible on the NVIDIA Jetsons to do a simple install from a RealSense debian repository (i.e. apt-get install). Previous versions of this repository require building librealsense from source, and (possibly) rebuilding the Linux kernel.
The current recommendation from Intel is to use UVC for video input on the Jetson family. The UVC API in librealsense has been rewritten to better support this use case.
$ ./installLibrealsense.sh
Note: You do not have to patch modules and kernels.
$ ./buildLibrealsense.sh [ -v | --version <version> ] [ -j | -jobs <number of jobs> ] [ -n | --no_cuda ]
Where:
<version>
= Librealsense version. E.g. v2.49.0<number of jobs>
= # of jobs to run concurrently when building. Defaults to 1 if the Jetson has <= 4GB memory<no_cuda>
= Compile without CUDA support. Defaults to CUDA.The librealsense Github repository has good documentation for supporting more advanced modes for the RealSense sensors. Please see: installation_jetson.md The documentation covers different communication interfaces and how to explore different features, some of which may require recompiling kernel modules.
Note: The build uses libuvc. You will not have to rebuild the kernel or modules in order to use this build.
Switch CLI argument to build_with_cuda ; Build with CUDA takes a lot more time because CMake needs to be rebuilt. Default is to build without CUDA support
previous commit Add CLI argument build_no_cuda ; script defaults to build with CUDA.
D435i is not recognized by RealSense applications, but shows up in Cheese webcam viewer.
Add CUDACXX flag for compilation using CUDA
Add USE_CUDA flag to script (Defaults to YES)
If using only RealSense T265 camera, this is the only installation necessary
If using the T265, you probably don't need CUDA (still needs to be tested); Set USE_CUDA to false. Saves compilation time
During compilation, the script will run out of memory on the Nano You will need either to:
OR:
patchUbuntu will patch all of the needed modules for librealsense, build the modules, and then install the modules. The kernel Image is then built and installed in /boot/Image.
Note: If you are building from a USB or some other device than the SD card, you will need to copy the Image file to the /boot directory on the SD card.