Nvidia Xavier NX Third Party Dev Board by SEEED Studio

Hello,

I am working with a Zed2 stereo camera and I am unable to optimize the neural depth feature supported. Investigating this problem I produced the following error in a verbose file: “Optimizing objects_performance_2.2 \ 0.3%[> ] 13min 17s est. left [ZED][WARNING] 2: [utils.cpp::checkMemLimit::380] Error Code 2: Internal Error (Assertion upperBound != 0 failed. Unknown embedded device detected. Please update the table with the entry: {{1794, 6, 16}, 12653},)
[ZED][DEBUG] Build engine successfully!
Segmentation fault (core dumped)”

Where the unknown embedded device is (as I believe) the A206 development board by SEEED Studio. We replaced the original development board with this one as the other suffered from damage during its use in a robotics application.

It should be noted we have a second Nvidia Xavier NX development board which is not a third party and we are able to optimize the AI module from Stereolabs. I believe this issue is similar to the one seen here:

where an individual ran into similar issues with a Jetson Nano. Please let me know if there are any workarounds for this because from what I understand I would have to buy and entirely new development kit to get just the development board. (I believe SEEED is a part of the Jetson Ecosystem Partners).

hello sjallan,

please help me understand this.
are you saying you have a damaged carrier board, and you would like to purchase this board individually instead of a whole developer kit?

Hello JerryChang,

Sorry if my last message was unclear. The problem is with the functionality of the ZED2 camera by Stereolabs. One of the features (neural depth mode) is unavailable and I cannot optimize the neural network to use alongside the device (which is supplied by Stereolabs). Investigating this problem with the Stereolab’s developers it was stated that this must be a problem with the third party board I was using (the A206 from SEEED). To draw this conclusion I generated a verbose file while running a python script as above to get the error: “Unknown embedded device found” which is similar to the error produced by another user in the forum I referenced which had an error with ONYX2TRT. I was wondering if there is a workaround to continue to use this third party board to gain the functionality from the Zed camera as this board is from one of the Nvidia Ecosystem Partners. Please let me know if I am on the right track here.

hello sjallan,

would like to double confirm the carrier board you’re using. is it the carrier board for Nano or Xavier NX?
for example, Xavier-NX eMMC module + Nano B01 carrier board is not a supported use-case.

Hello JerryChang,

I am using the third party carrier board by SEEED Studio which is designed for the Nvidia Xavier NX.

hello sjallan,

please test ZED2 camera with V4L2 IOCTL to verify basic camera functionality,
you may see-also developer guide, Applications Using V4L2 IOCTL Directly.
please share the test results, and also kernel messages for reference,
thanks

Hello JerryChang,

I know this issue has been closed, but I realized I forgot to post the solution that worked for me. This Github issue references the same problem: Error Code 2: Internal Error (Assertion upperBound != 0 failed. Unknown embedded device detected. Please update the table with the entry · Issue #1866 · NVIDIA/TensorRT · GitHub and the patch released by Nvidia here: Jetson Linux R32.7.2 Release Page | NVIDIA Developer solved the " Unknown embedded device detected" error I have receiving. The patch is: TensorRT_8.2.1.9_Patch_for_Jetpack4.6_Jetson_NX_16GB.tar.gz