Can't build OpenNI2 on NVidia Jetson TX2


#1

I’m trying to build OpenNI2 on the NVidia Jetson TX2 (aarch64 platform) by following a combination of instructions in the github repo, and this article. I’m running into loads of problems actually making the binary, though.

When I run make, I get the following error: “ThirdParty/PSCommon/BuildSystem/CommonDefs.mak:22: *** Can’t determine host platform. Stop.”

So, taking a risk that the aarch64 platform is “close enough” to the “arm” platform, I hardcoded that into CommonDefs.mak. Then I get the errors:

g++: error: unrecognized command line option ‘-mfpu=neon-vfpv4’
g++: error: unrecognized command line option ‘-mfloat-abi=hard’

Does anyone have a way to build OpenNI2 to run on a Jetson TX1 or TX2 (Ubuntu 16.04)?


#2

Unfortunately, I do not have a TX2 in the office to test out the integration, but I thought I would send this over to you to see if it helps:

Don’t know if you have see that page yet or not!


#3

Thanks. That’s basically where I started, but it doesn’t work on TX2 using the latest L4T. This has actually been fixed in a pull request (https://github.com/occipital/OpenNI2/pull/116 ), but since the repo doesn’t seem to be maintained anymore, I ended up just cloning someone’s fork and building that instead: https://github.com/mikeh9/OpenNI2-TX1 .


#4

I just found this, as well, which might be useful to you and other users: https://github.com/EllieOwen/3DscanUbuntu/tree/master


#5

Doesn’t seem to support aarch64 at all.


#6

For other readers – Check Philippe’s solution at very bottom of link in OP