Thank you JerryChange.
Can explain why imx219 board setup error ( [ 2.022231] imx219 9-0010: imx219_board_setup: error during i2c read probe (-121) ) is image quality issue. I feel it means that the board for imx219 is not ready. And there is only one file under /var/nvidia/nvcam/settings. The file is README.txt
Thank you JerryChange.
I am of the same mind, it is nVidia’s proprietary chip, proprietary drivers that have changed. The cameras are the same Rasp Pi V2’s, and do not come with vendor supplied drivers. They are a defacto standard and advertised to work.
As far as I am concerned the device is defective.
The moderator JerryChang’s advice is of no value. I have the spec sheet for the sensors (from Sony), there is no method for me to enter them and infromation on the tegra chip is not available.
So the device does not meet the purpose for which it was bought.
I might just email the original developers and see if they will respond.
since you’re able to enable Raspberry Pi v2 cameras (IMX219) on NX platforms, it means sensor driver and camera functionality works. please initial another discussion thread if you got issue to enable sensor on NX.
please refer to Camera Architecture Stack for the [Tuning] block. there’s further step for IQ, there’re tuning parameters that should direct to sensor vendors. you should also note that same calibration file cannot apply to different tegra family. every sensors doing the same, it need another image calibration process to generate appropriate tuning parameters.
may I know your sensor vendors, please list the details for reference.
NO, you’re bump into sensor bring-up issue, please initial another discussion thread for tracking.
The camera I am using is a genuine Raspberry Pi V2. nVidia claims it is supported. Please contact your internal folks and find out what you can do to help. The Camera Vendors will not! They support only Raspberry Pi.
The problem I have is:
- large white Dots in fixed position on the image,
- Excessive noise and blurred images.
This paragraph taken from Page 5 of the nVidia Xavier NX Developer Kit EXPLCITLY!!! states that the NX supports this camera.
[J1] Camera connector; enables use of CSI cameras. The Jetson Xavier NX Developer Kit works with IMX219 camera modules, including Leopard Imaging LI-IMX219- MIPI-FF-NANO camera module and Raspberry Pi Camera Module V2.
(Page 5 from the attached official nVidia document)
For reference, this is your companies documentation!
it seems you still don’t understand what’s the process there; your camera streaming works, isn’t it? it always an extra effort for [Tuning] according to the [Camera Architecture Stack].
please check the attach binary Topic139727_Jul09.zip (780.4 KB) which include some tuning parameters as an example,
you should replace /usr/lib/aarch64-linux-gnu/tegra/libnvodm_imager.so binary with attachment. you should also perform warm-reboot to take effect.
as I mentioned previously, we don’t provide image tuning support via forum discussion threads.
IQ tuning depends-on your hardware lens cover and also your actual use-case.
please, you should contact with sensor vendor to explain what you actually need is IQ tuning.
I have the same white pixels issue with IMX577 sensor and NX. I have captured yuv 4k stream. White pixels were added to the image. Is there a fix to this issue? The ISP should remove those pixels?
YUV frame is attached
Beware! nVidia is “puffing” (marketing term for lying) when they say the raspberry pi cameras are supported.
It turns out you need to “tune” the camera to get it to work properly. nVidia will give you useless advice because you need deep knowledge on how to execute his elitist suggestions / help. He provides us with NO advice on how to tune the camera yourself and brushed off the problem by supplying the “binary file” in the post above. … Nice file with cute bytes but a waste of electrons without documentation on what to do with it.
Support from the camera manufacturer as suggested by nVidia is again useless advice, he clearly does not know what support exists for the “nVidia supported” raspberry pi camera. Support is NOT available from camera vendors supplying “genuine” raspberry pi cameras because they only support the raspberry pi which makes sense. What is worse, the “nvarguscamerasrc” drivers are all written by nVidia. There is no documentation I could find on how tune, modify or configure these. They do not seem to be actively supported by nVidia. So basically unless you use the Jetson Nano you are out of luck with Raspi cameras.
Bottom line, you need to spend another $100(us) or more and buy a proprietary camera from an nVidia partner to get the support that nVidia says we need. I bought one from India. It was $135 with shipping. I will post here once I get results. There is also one from ArduCam. Both of these cameras come with a dedicated circuit board to interface with the NX. I have no idea how compatible they will be, so time will tell.
You can try lighting the subject with a bright light it may help a bit.
may I know had you ever tried the pre-built library in post #28, please attach the capture results.
please, you should study what’s the “tuning” actual means.
you may also share your real use-case, you should contact with nvidia sales team if you would like dedicated tuning supports.
as I mentioned previously, we don’t provide image tuning support via forum discussion threads,
camera image tuning always specific to each modules;
that pre-built library in post #28 for updates for NX/ Raspberry Pi V2, IMX219
please contact with your sensor vendor for tuning supports,
please note that,
you MUST have image tuning process for each bayer sensors, a.k.a MIPI-CSI sensors;
you may choose USB sensors, which bypass internal ISP for plug and play camera solution.
please check Jetson Partner Supported Cameras,
there’s a list of cameras supported by Jetson Camera Partners on the Jetson platform.
I have no idea how to do that Jerry. So no I have not used it. Please point me in the correct direction to how to do this.
Please provide instruction on how to use the pre-built library.
you may download the attach binary, Topic139727_Jul09.zip,
extract that and you’ll see Topic139727_Jul09_libnvodm_imager.so which include some tuning parameters as an example.
copy that pre-built to your target, replace it with /usr/lib/aarch64-linux-gnu/tegra/libnvodm_imager.so;
$ sudo cp Topic139727_Jul09_libnvodm_imager.so /usr/lib/aarch64-linux-gnu/tegra/libnvodm_imager.so
you should also perform warm-reboot, i.e.
$ sudo reboot, to let the changes take effect.
please have confirmation,
you might also consider to modify post #31 to revise those incorrect info.
Jerry… please stop saying “contact the camera or sensor vendor” they do NOT support nVidia only raspberry pi. You are wasting forum members time when they try to get this support!!!
All you need to say is: nVidia is sorry. We were incorrect in stating support for the raspberry pi camera on the Xavier NX this was an error.
Please purchase a camera from an nvidia partner to get camera support.
These partners can provide IMX219 support via a proprietary ISP card.
again, had you tried to download the attach binary, Topic139727_Jul09.zip for verification?
I will try this is the morning.
Not sure what info is incorrect but I will review it. Please be specific on your concerns.
I will try it in the morning. It is late here now.
there’re two steps, (1) camera functionality and (2) image quality (IQ) tuning.
since you can plug Raspberry Pi V2 to NX platform and confirm preview stream works, then it’s already supported.
it always an extra effort for IQ tuning, it depends-on your actual use-case and also IQ expectation.
I’ve already share pre-built binary in post #28 to include some tuning parameters as an example, as I mentioned several times, we don’t provide image tuning support via forum discussion threads.
there’s correct path for asking IQ tuning,
please contact with nvidia sales team if you would like dedicated tuning supports.
Regarding the ISP pipeline. I do not know if it will ever happen, but INMO you should open and give more control to the ISP pipeline for developers. I don’t have the statistics but I assume that the main use-case when using Jetson/Xavier is for vision applications. In this case Image Quality is very important. When ‘outsourcing’ Image Quality tuning for 3rd parties that cause for longer support response time for customers and a lot of frustration. This is a big disadvantage that obscure the other advantages of Jetson/Xavier advantages.
there’re Jetson Camera Partners, they’re also provide image quality tuning supports.
or, if you would like to have dedicated image tuning supports, please contact with nvidia sales team.