Leopard Imaging LI-AR0231-GMSL USB Camera + GStreamer on TX2

I have connected a Leopard Imaging LI-AR0231-GMSL 1080p HDR GMSL Serial Camera to my TX2 via the LeopardImaging USB3 GSML Deserializer. The camera outputs 12bit RCCB RAW data. Datasheet:

The camera appears as /dev/video1 - and I can show the image running this GStreamer command.

gst-launch-1.0 v4l2src device=“/dev/video1” ! 'video/x-raw, format=(string)I420’ ! nvoverlaysink

The image is displayed like this:

I have tried with different parameters like video/x-bayer, bayer2rbg etc - but to no avail.

Is the problem that RAW12 is not supported?
Could anybody shed some light here and point me in the right direction?

(I would like bayer conversion to be done by the ISP’s - in order to keep latency as low as possible.)

this is leopard team.
as far as we know currently the internal isp do not support usb cam. you need convert the raw data to yuv format manually.
you can reference guvcview code.


Isn it possible to:

  • Write a UVC Driver which converts the raw image using libArgus (which uses the ISP) ?
  • Write a GStreamer plugin that uses libArgus to do the conversion ?


Hi njal,
Do you connect the camera to USB 3.0 type A port on developer kit?

Hi Danell

We gave up on Leopard Imaging alltogether. Could not get it to work on either Linux or Windows.

We used USB3 cameras from https://www.theimagingsource.com instead. The installation took like 1 minute, and the cameras has also has a blazing fast onboard FPGA that does debayering etc.


are there any updates on this? We have the same problem with our LI-USB30-AR0231-GMSL-R7 Cam. Cannot generate any clear image, neither on Windows nor on Linux or TX2. Is there any sample code LI could provide?

Amazed that Leopard Imaging sells such products without any clear documentation or samples…

No updates here. Steer clear of leopard imaging. Support is non existent.

Hey, we have been working with AR0231 and GMSL too and we were finally able to get it working, in case you need help, you can read more about it here:

We are using a Sekonix camera (http://sekolab.com/products/camera/).