Too slow initialization when using gstreamer in docker

[choi_giryong]

Hi.
I am trying to get RGB data played by camera.
My test gstreamer command is this:

gst-launch-1.0 nvv4l2camerasrc device=/dev/video0 name=vid0 ! “video/x-raw(memory:NVMM), width=1920, height=1536, format=UYVY, framerate=30/1” ! nvvideoconvert compute-hw=2 nvbuf-memory-type=4 ! “video/x-raw(memory:NVMM), format=(string)I420, width=(int)1920, height=(int)1536” ! nvvideoconvert compute-hw=1 nvbuf-memory-type=1 ! “video/x-raw(memory:NVMM), format=(string)RGB” ! nv3dsink sync=false

This works fine in host, but it takes too long time to initialize (about 1~2 minutes) in docker.
However, once initialized, it plays normally although in docker.
I noticed that this command works well:

gst-launch-1.0 nvv4l2camerasrc device=/dev/video0 name=vid0 ! “video/x-raw(memory:NVMM), width=1920, height=1536, format=UYVY, framerate=30/1” ! nvvideoconvert compute-hw=2 nvbuf-memory-type=4 ! “video/x-raw(memory:NVMM), format=(string)I420, width=(int)1920, height=(int)1536” ! nvvideoconvert ! “video/x-raw(memory:NVMM), format=(string)NV12” ! nv3dsink sync=false

I think this is because second “nvvideoconvert” in first command uses GPU not VIC, but not sure, and I need RGB output.
Is there any solution to solve this?

[carolyuu]

Hi,

For the camera basic functionality first needs to check the device and driver configuration.
You can reference to below program guide for the detailed information of device tree and driver implementation.
https://docs.nvidia.com/jetson/archives/r36.3/DeveloperGuide/SD/CameraDevelopment/SensorSoftwareDriverProgramming.html?highlight=programing#sensor-software-driver-programming

Please refer to Applications Using V4L2 IOCTL Directly by using V4L2 IOCTL to verify basic camera functionality.
https://docs.nvidia.com/jetson/archives/r36.3/DeveloperGuide/SD/CameraDevelopment/SensorSoftwareDriverProgramming.html?highlight=programing#to-run-a-v4l2-ctl-test

Once confirm the configure and still failed below link help to get log and some information and some tips for debug.
https://elinux.org/Jetson/l4t/Camera_BringUp#Steps_to_enable_more_debug_messages

Thanks!

[DaneLLL]

Hi,
Please share your release version. The latest release is Jetpack 6.1. If you use previous version, would suggest upgrade.

And please check

1. Does it occur outside docker?
2. In docker, does it occur if you use VIC to convert to RGBA?

[choi_giryong]

Jetpack version is 5.1.2.
It is hard to change version because camera supplier doesn’t support latest version yet.

1. Does it occur outside docker?
   No. It is only occur in docker.
2. In docker, does it occur if you use VIC to convert to RGBA?
   No. When only I use GPU the problem occur.

[DaneLLL]

Hi,
We will check it on Jetpack 6.1. Since you use 5.1.2, you may consider not to use nvvideoconvert in the use-case. VIC supports 32-bit RGBA and BGRx. May check if you can use the VIC-supported formats in the use-case.
Or you may consider use jetson_multimedia_api. It is more flexible and you can implement CUDA code to re-sample RGBA NVvBufSurface into RGB CUDA buffer.