@Irakandjii, it should be simply replace a pre-built binary for verification, please attach your capture results. thanks
I was able to install the .so file.
There was an improvement to the image, the number of white dots is significantly reduced.
During my most recent research, I noted the following fact in the Jetpack 4.4 Release notes.
Noisy corruption with Preview/Capture observed on Jetson Xavier NX with IMX219.
Tuning is in progress.
So this has been a known issue for quite a while and nVidia is doing the tuning to fix it.
I could have saved hours of time and avoided significant expenditure if you had informed this forum that there was an active Issue and the problem would be resolved in a future update.
So it appears nVidia does support the Camera.
I am confused why this forum is not up to date on the internal “reported issues” list.
This disparity has resulted in inaccurate advice.
Reading again the whole thread to can believe it. Unreal and unacceptable. Without enter the money, time, project damage and frustration that this could made.
@Irakandjii, thanks for confirmation.
it’s Top Fixed Issues: 200585128, which means l4t-r32.4.3 already include the fixes;
please noticed that IMX219 (Raspberry Pi version) has unusually high noise level with bad pixels, especially in the low-light environment; that’s why this topic was created.
we had deliver an update in post #28, please apply that manually based-on JetPack-4.4 / l4t-r32.4.3;
we had also complete the code-review process, you may expect next public release version (i.e. l4t-r32.5) will include the fixes.
I have applied the update from post #28 and the results are still very poor compared with the Jetson Nano. I have the latest release and there doesn’t seem to be a significant change when replacing the libnvodm_imager.so with the version in post #28?
In low light conditions the white pixels are still very extreme and with the same camera parameters they are almost non existent on the Jetson Nano???
I understand that this is a poor sensor, but will there be an update to provide the same low light performance on the Xavier Nx as the Jetson Nano when using the IMX219?
I have also tested the Jetson Nano libnvodm_imager.so on the Xavier NX and it removes the noise; however the image is left looking a bit washed out. I can provide image comparisons with each file version and the results from the Nano in a later post if that is of use?
It certainly seems like the tuning has not been done even in the latest release and it is very crippling to continue stating that the camera is supported (I know the process, so please don’t repeat that again). Since this is likely to be a widely used camera for testing and evaluation of the platform please take the time to properly support the IMX219 raspberry pi v2 camera as you do with the Jetson Nano.
To me the issues are still not resolved in the latest release and I am very disappointed to see that NVIDIA has stated that the issue has been fixed when it appears to not be the case???
I look forward to the release of l4t-r32.5 and hold hope that this issue has been resolved. In the mean time do you have an updated version of libnvodm_imager.so that you can share? Or is the version from post #28 the same one that is in the upcoming release l4t-r32.5 ?
Both of the following images were captured on the Jetson Xavier NX with the raspberry pi v2 camera in cam 0 port. Three versions of the libnvodm_imager.so file were used, however I am a new user and can only upload two images and post three replies.
The image below was captured with the Topic139727_Jul09_libnvodm_imager.so file:
The image using the NX orignal file looks much the same, but the image in the following reply looks much better using the Jetson nano libnvodm_imager.so file.
Any advice on the future of tuning parameters for this camera on the Xavier NX is appreciated.
could you please also compare the image quality without and with the patch in post #28.
since this is already one long discussion thread, and we had an update in post #28. you may expect next public release version (i.e. l4t-r32.5) will have same fixes.
this should be a follow-up inquiries,
I would suggest you initial another new discussion thread for better supports.
Im having the same problem as discussed above, i have tried all the suggestions and alot of parameters tuning with no progress, but with the same piece of code on my nano the image is perfctly clear.
I paid 4X the price of nano and the quality is nowhere near.
do you have and new solution? a different approach? please check it yourself and i will be more than happy if youll share the sulotion.
Actually I gave up. I purchased a Hi res camera from Arducam and got it working with their drivers, It is a hi-resolution 477. I found image was ok, but I took a performance hit.
I also purchased a camera from e-con, but they are very bad at support. You get an initial driver but have to pay for ongoing support. It was a waste of money.
Overall I spent more on cameras than on the NX and never really got it working.
I have not turned the NX on since July 2020. I might try again now there is a new Jetpak.