How to Build Docker Container with GStreamer-enabled OpenCV for Jetson Platform?

[shaunliew]

Hi @dusty-nv,

I'm working on implementing YOLOv11 with a CSI camera in the Ultralytics Docker container on my Jetson Orin Nano running JetPack 6.1. I notice that your jetson-inference Docker container has OpenCV built with GStreamer support, while the Ultralytics container does not.

Current Setup:

• Jetson Orin Nano Super Kit (JetPack 6.1 and L4T 36.4.2)
• CSI camera (IMX219 Raspberry Pi Camera)
• ultralytics/ultralytics:latest-jetson-jetpack6 Docker image
• Following your container setup guide with proper device mounts

Issue Description:
I'm unable to access my CSI camera from within the Ultralytics Docker container, but the camera works perfectly on the host system. When using your jetson-inference container setup, everything works correctly. After investigation, I found that the root cause is that OpenCV in the Ultralytics container is built without GStreamer support.

I've mounted all necessary devices and paths:

sudo docker run -it --rm \
    --runtime nvidia \
    --network host \
    --ipc=host \
    -v /tmp/argus_socket:/tmp/argus_socket \
    -v /etc/enctune.conf:/etc/enctune.conf \
    -v /etc/nv_tegra_release:/etc/nv_tegra_release \
    -v /tmp/nv_jetson_model:/tmp/nv_jetson_model \
    -v /var/run/dbus:/var/run/dbus \
    -v /var/run/avahi-daemon/socket:/var/run/avahi-daemon/socket \
    --privileged \
    $DISPLAY_DEVICE $V4L2_DEVICES \
    $t

I've verified that:

• Camera is visible at /dev/video0 (confirmed with v4l2-ctl --list-devices)
• GStreamer works in the container (nvgstcapture-1.0 --orientation 2 runs successfully)
• All necessary devices are mounted properly

However, when trying to access the camera through OpenCV using GStreamer pipeline:

video_capture = cv2.VideoCapture(gstreamer_pipeline(flip_method=0), cv2.CAP_GSTREAMER)

I get Error: Unable to open camera

Root Cause Identified:
The OpenCV in the Ultralytics container lacks GStreamer support, while your jetson-inference container has it enabled:

Your jetson-inference container shows GStreamer support:

Video I/O:
    DC1394:                      NO
    FFMPEG:                      YES
      avcodec:                   YES (58.134.100)
      avformat:                  YES (58.76.100)
      avutil:                    YES (56.70.100)
      swscale:                   YES (5.9.100)
      avresample:                NO
    GStreamer:                   YES (1.20.3)
    v4l/v4l2:                    YES (linux/videodev2.h)

Ultralytics container shows no GStreamer support:

Video I/O:
    DC1394:                      NO
    FFMPEG:                      YES
      avcodec:                   YES (59.37.100)
      avformat:                  YES (59.27.100)
      avutil:                    YES (57.28.100)
      swscale:                   YES (6.7.100)
      avresample:                NO
    GStreamer:                   NO
    v4l/v4l2:                    YES (linux/videodev2.h)

Questions:

1. Could you share how you configured OpenCV build in your Dockerfile to enable GStreamer support?
2. Are there specific dependencies or build flags I should include in my Dockerfile?
3. Do you have any recommendations for building OpenCV with GStreamer support specifically for Jetson platforms?

The goal is to replicate your successful GStreamer-enabled OpenCV setup in my Ultralytics container for YOLOv11 development. Any guidance would be greatly appreciated! Thank you for maintaining this excellent repository.