Raspberry Pi HQ Camera and Xavier

Hello Jerry,

according to the Jetson Nano thread the HQ camera does not work. The plugin is nvarguscamerasrc. So I think it will not work with Xavier NX too.
Best regards,
Wilhelm

hello WiSi-Testpilot,

Raspberry Pi HQ Camera was using Sony IMX477R 12.3-megapixels sensor.

we did not maintain IMX477R sensor drivers, you may check L4T Sources for all supported camera sensor drivers.
for example,
$L4T_Sources/r32.4.2/Linux_for_Tegra/source/public/kernel/nvidia/drivers/media/i2c/*

BTW, you might also contact with Jetson Preferred Partners for camera solutions.
thanks

Hello,

Currently you have support for the IMX219 Raspberry cameras, is there a plan for implementing the IMX477?

hello fredrik.mattsson,

NO. currently do not have plan to implement IMX477 as reference drivers.
you should contact with Jetson Preferred Partners for asking camera solutions.
thanks

So, raspiraw is a program/codebase for retrieving raw sensor data from the camera on raspberry pi. This file has some structs for register addresses and values for different modes for the hq camera. These are “closed source” so they dont document or tell what the registers and values correspond too, but perhaps they could be used as is for some initial POC drivers? File here:

I saw this, in more of a place, RPi to HQ Quality camera is compatible with Xavier Nx
I bought the Nx, camera HQ and a special fisheye lens ,a nvme pci ssd, and my project is based on this components, please tell that those will work together

there is a driver for IMX477 by Leopard Imaging for Jetson;
However, it was never tested if it works with the Rpi.high.res camera. Maybe it will require too difficult customization to work with it.
There is also an option to order development of the driver for Rpi.high.res camera from commercial development teams[ e.g. RidgeRun]

Yeah, bit I put my focus in what a official dialer as Seeed Studio and a press website as Tom’s Hardware said it is perfectly compatible and I bought a lot of things based in that information.

if you contact mentioned sources you might be able to get them either to correct information at the web pages or to provide the extended information regarding the compatibility, as they meant it

Seeed Studio and a press website as Tom’s Hardware

In my opinion Tom’s Hardware pointed to Raspbery Pi v2 camera stating that it is compatible with nano/nx;
On the other hand, seed studio might have had a glitch with their database that could cause wrong categorization of the device

Thanks by the answer, I really dont think is a error, they are using it as a good point to sell the Xavier, and indeed worked with me.

They mention clearly the “new” “just released” camera.
I don’t know from where they took that info, a nVidea press note or dossier maybe? They invented?

Hi All,

RidgeRun is currently working on the Jetson driver for the new HQ Raspberry Pi camera. We hope to get it ready in the incoming weeks. You can read more about it here:

-David

1 Like

That are great news, I just want give you thanks by share with us, without people or companies like you/yours, beginners, people with limited resources as me could not have access to so many things.
Sincerely thanks (gracias)

Hi

Would ask to reconsider that decision about not supporting the IMX477 as reference drivers, as these cameras do provide quite a good quality for a very reasonable price range. In my case I wanted to have 2 of these connected (with different lenses) at the same time together with a Intel RealSense to enable navigation for drones in GPS denied environments, and the regular RPI cams don’t provide enough image quality for the goal.

Best

hello luisvale,

it’s not simply adding reference drivers,
you may also access L4T sources release package for the kernel sources, checking reference camera sensor drivers as below.
for example,
$L4T_Sources/r32.4.2/Linux_for_Tegra/source/public/kernel/nvidia/drivers/media/i2c/

we had QA testing farm with these sensor modules to validate camera software stack,
please contact with sales team to clarify business requirement, or, you should contact with Jetson Preferred Partners for camera solutions.
thanks

Thanks

I’m not a kernel dev, I expressed my opinion, which I would appreciate could be transmitted up to the kernel dev team to consider adding this module to the supported list.

1 Like

Hi David
It looks you updated the wiki with further info of the module compatibility, if I make that hardware modification the camera will work? The use will be just like the rpi camera v2?

Thanks

Hi fpsychosis,

With this change you will be able to see the camera in the i2c bus. This is mandatory to let the driver configure the camera and for the Jetson to capture. However, we are still in the process of testing the driver. You can read more about it here. I recommend you to wait on us to release the driver before making the hardware change. You won’t be able to capture without the driver. We are working on it to release it as soon as possible.

-David

1 Like

Crystal clear. To wait then, thank you!

Hi David,
thank you for your work. I am very excited.
It would be very nice if the behaviour of the HQ is the same as the R-Pi V2 Camera.
I mean, that at low illumination the camera automatically increases the gain.
(On the R-Pi with raspivid it is simply dark.)
Then it is possible to accumulate frames to reduce the S/N and to get picture, which are below the noise, see my simple example (picture and code).
Best regards,
Wilhelm

import numpy as np
import cv2
import time

def willi():

    cap = cv2.VideoCapture("nvarguscamerasrc maxperf=true ! video/x-raw(memory:NVMM), width=1280, height=720, format=(string)NV12, framerate=60/1 ! nvvidconv flip-method=2 ! video/x-raw, width=1280, height=720, format=(string)BGRx ! videoconvert ! video/x-raw, format=(string)BGR ! appsink", cv2.CAP_GSTREAMER)
    if cap.isOpened():
        cv2.namedWindow("willi", cv2.WINDOW_AUTOSIZE)
        counter = 0
        while cv2.getWindowProperty("willi", 0) >= 0:

            ret, csi = cap.read()
                    
            cv2.imshow("CSI Kamera", csi)
      #     grau = cv2.cvtColor(csi, cv2.COLOR_BGR2GRAY)
      #     cv2.imshow("CSI in s/w", grau)

            summe = np.asfarray(csi) 
      #     summe = csi.astype(int)
            summe = summe * 0.0

            for i in range (0,100):
                ret, csi = cap.read()
    #           summe = summe + csi.astype(int)        
                summe = summe + np.asfarray(csi)
   #            summe = summe + ftemp

            summe = summe / (255 * 100)
            cv2.imshow("Average", summe )
            counter = counter + 1
            print("Sekunde: ", time.strftime("%S"),"100 x Acc:",counter)

            keyCode = cv2.waitKey(30) & 0xFF
            # Stop the program on the ESC key
            if keyCode == 27:
                break

        cap.release()
        cv2.destroyAllWindows()
    else:
        print("Unable to open camera")

if __name__ == "__main__": willi()