Connecting RGB888 format camera device to TX2

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hello all,

I am trying to input RGB888 image data from Camera device to TX2, but it does not work. I referred several forum topics, and I understood as follows.

· JETSON’s V4L2 does not support RGB888 format, so mapping to AR24 (BGRA) is recommended.
<https://devtalk.nvidia.com/default/topic/1042552/jetson-tx2/mipi-signal-from -fpga-on-jetson-tx2-evaluation-board/>
I tried to create Image file corresponding to input RGB888, but pixel format was converted to AR24 automatically at V4L2. And, acquisition of RGB888 data failed.

· On the other hand, NVCSI supports RGB888, but not supports AR24 (TRM; Input Data Format 27.10.1).
https://devtalk.nvidia.com/default/topic/1036742/jetson-tx2/tx2-camera-not-set-rgb888/
https://devtalk.nvidia.com/default/topic/1042634/jetson-tx2/nvcsi-vi-pixel-format-doubts/
TX2 does not start up by using Image file corresponding to AR24. Is this because NVCSI is not support AR24?

It looks there is no way to connect a device that outputs RGB888 to TX2.
If there is a way, I would like to know it.

Thank you.

2

hello sumisumisumi,

we don’t support RGB888 by default,
please refer to /kernel_src/kernel/kernel-4.4/include/uapi/linux/videodev2.h
you might have a try to configure it as below.

#define V4L2_PIX_FMT_RGB24   v4l2_fourcc('R', 'G', 'B', '3') /* 24  RGB-8-8-8     */

you should also extend the supported pixel formats as below.
/kernel/kernel-4.4/drivers/media/platform/tegra/camera/sensor_common.c

static int extract_pixel_format(const char *pixel_t, u32 *format)
{
...
}

/kernel/kernel-4.4/drivers/media/platform/tegra/camera/camera_common.c

static const struct camera_common_colorfmt camera_common_color_fmts[] = {
...
}
3

Hello, JerryChang

Thank you for quick response.
However, I already tried pixel formats extension code as follows.

<sensor_common.c>
static int extract_pixel_format(const char *pixel_t, u32 *format)

else if (strncmp(pixel_t, "rgb", size) == 0)
        *format = V4L2_PIX_FMT_RGB24

<camera_common.c>
static const struct camera_common_colorfmt camera_common_color_fmts =

{
  MEDIA_BUS_FMT_RGB888_1X24,
  V4L2_COLORSPACE_SRGB,
  V4L2_PIX_FMT_RGB24,
}

However, V4L2 automatically converts it to AR24.
Also, I tried accessing data as AR24. But I could not acquire any data.
Does it need more additional process?

When I tried the same thing in UYVY format, it worked without problems.

4

hello sumisumisumi,

could you please share details logs about the failure.
also, may I know how did you consider that V4L2 automatically converts it to AR24.
thanks

5

sumisumisumi

We were able to get an rgb888 sensor working on Tx2 and Xavier. Other than adding code similar to what has been mentioned in this thread already I don’t think we needed to anything special to get the camera to work.

We used this basic pipeline for verification:

gst-launch-1.0 v4l2src ! videoconvert ! ximagesink
6

JerryChang
D3_growe

I tried executing the following command, but it was not working.

gst-launch-1.0 v4l2src ! videoconvert ! ximagesink

message is as follows.

Setting pipeline to PAUSED ...
Pipeline is live and does not need PREROLL ...
Setting pipeline to PLAYING ...
New clock: GstSystemClock

Processing stops here.
gst-launch-1.0 icon appears, but the picture is not displayed.
When I close icon, error message appears as shown below.

ERROR: from element /GstPipeline:pipeline0/GstXImageSink:ximagesink0: Output window was closed
Additional debug info:
ximagesink.c(700): gst_x_image_sink_handle_xevents (): /GstPieline:pipeline0/GstXImageSink:ximagesink0
Execution ended after .......
Setting pipeline to PAUSED ...
Setting pipeline to READY ...
Setting pipeline to NULL ...
Freeing pipeline ..

Also, if I use the following command, the Pixel Format becomes ‘AR24’ even though it is set as RGB888.

v4l2-ctl --list-formats

Output is

ioctl: VIDIOC_ENUM_FMT
     index       : 0
     Type        : Video Capture
     Pixel Format: 'AR24'
     Name        : 32-bit BGRA 8-8-8-8

Is it okay to ignore this?

Thank you

7

How to make this RGB888 instead of ‘AR24’ ? Could you do it?

8

hello hburaksaruhan,

let’s tracking at Topic 1051967 since you initial another discussion thread for this, thanks