Device CloudQualcomm® Device Cloud

RTSP Streaming

Qualcomm Dragonwing™ RB3 Gen 2 RTSP Camera Streaming with Qualcomm Linux 1.3 AI Sample Application Demo on Qualcomm Device Cloud

How to do RTSP camera streaming on the Qualcomm Dragonwing™ RB3 Gen 2 devices?

The Qualcomm Linux AI sample applications can take camera, file, and RTSP stream input sources and outputs. Since you will be using a remote RB3 Gen 2 in Qualcomm Device Cloud, RTSP camera streaming will allow you to stream a live camera input from your local machine (e.g. a webcam, attached USB camera, or local network RTSP server) and output to a RTSP camera stream for playback on your local machine.

Qualcomm Dragonwing™ RB3 Gen 2 RTSP Camera Streaming Diagram over QDC

Prerequisites

  • Qualcomm Dragonwing™ RB3 Gen 2 device with Qualcomm Linux 1.3+ and QIMSDK
  • FFmpeg: For capturing attached cameras into an output RTSP stream.
    • Camera capture: Windows DirectShow, video4linux2, and OS X AVFoundation
    • The OS specific ffmpeg documentation also provides commands to find the camera name and pixel format depending on the camera hardware.
  • mediamtx: A simple RTSP server for proxying the local machine camera stream.

Instructions

  1. Start a QDC Interactive Session with a RB3 Gen 2 device.

    • Upload or choose an existing SSH public key for the interactive session to enable SSHing to the RB3 Gen 2 device.

  2. In the running interactive session, click on the Connect button to view the SSH instructions to connect to the device.

    • The additional ports mentioned in the last section of the instructions will be needed for RTSP camera streaming.
    • Connect to the RB3 Gen 2 device with SSH. For example:
    # Port 5037 is the ADB server
    # Port 8900 will be used for the mediamtx RTSP server
    # Port 8901 will be used by the GStreamer RTSP server
    # Port 22 is the RB3 Gen 2's SSH server which will be reached via localhost port 2222
    

    ssh -i <PRIVATE_KEY_FILE_PATH> -L 5037:sa<session>.sa.svc.cluster.local:5037 -L 2222:sa<session>.sa.svc.cluster.local:22 -L 8900:sa<session>.sa.svc.cluster.local:8900 -L 8901:sa<session>.sa.svc.cluster.local:8901 -N sshtunnel@ssh.qdc.qualcomm.com

  3. Check if you can SSH to the device. Alternatively, you may use adb shell and adb push if preferred. ssh and scp has replaced adb in the Qualcomm Linux 1.3 documentation.

    ssh -i <PRIVATE_KEY_FILE_PATH> -p 2222 -o StrictHostKeychecking=no -o UserKnownHostsFile=/dev/null root@localhost

  4. Copy the mediamtx tarball file downloaded from the prerequisites section to the device:

    # e.g. mediamtx_v1.11.3_linux_arm64v8.tar.gz
    scp -i <PRIVATE_KEY_FILE_PATH> -P 2222 -o StrictHostKeychecking=no -o UserKnownHostsFile=/dev/null <local_file_location>/mediamtx_<version>_linux_arm64v8.tar.gz root@localhost:/opt/mediamtx.tar.gz

    Run the following in ssh:

    # In an ssh session
    cd /opt
    tar -xf /opt/mediamtx.tar.gz

    Edit the mediamtx.yml file so the RTSP server port is one of the opened output ports you will be using (i.e. 8900-8909):

    # In an ssh session
    vi /opt/mediamtx.yml

    mediamtx.yml snippet example:

    ###############################################
    # Global settings -> RTSP server
    

    # Enable publishing and reading streams with the RTSP protocol.
    rtsp: yes
    # List of enabled RTSP transport protocols.
    

    ...
    

    # Address of the TCP/RTSP listener. This is needed only when encryption is "no"
    rtspAddress: :8554

    Change the rtspAddress: :8554 portion:

    # Address of the TCP/RTSP listener. This is needed only when encryption is "no"
    rtspAddress: :8901

    Alternatively, you can edit the file on your local machine before pushing to the RB3 Gen 2 device.

  5. Export and push your YOLOv5 model generated from the Qualcomm Linux 1.3 documentation for the Multi-input AI inferencing sample application.

    scp -i <PRIVATE_KEY_FILE_PATH> -P 2222 -o StrictHostKeychecking=no -o UserKnownHostsFile=/dev/null yolov5.labels root@localhost:/opt/yolov5.labels
    scp -i <PRIVATE_KEY_FILE_PATH> -P 2222 -o StrictHostKeychecking=no -o UserKnownHostsFile=/dev/null yolov5.tflite root@localhost:/opt/yolov5.tflite

  6. Start the mediamtx RTSP server on the RB3 Gen 2.

    # Run in an ssh session
    cd /opt
    ./mediamtx

  7. Use FFmpeg to send a local RTSP camera stream to the mediamtx RTSP server. See prerequisites for details for running FFmpeg camera capture on other operating systems. For example, on a Windows laptop with a webcam named "Integrated Camera":

    ffmpeg -f dshow -video_size 640x480 -pixel_format yuyv422 -i video="Integrated Camera" -framerate 15 -tune zerolatency -vf format=yuv420p -c:v h264 -crf 10 -f rtsp -muxdelay 0 -rtsp_transport tcp rtsp://127.0.0.1:8901/mystream/live1.mkv

    See prerequisites for details for running FFmpeg camera capture on other operating systems or how to list available camera sources. Note: The "live1.mkv" part of the URL is expected from the gst-ai-multi-input-output-object-detection sample application.

  8. Create the weston.ini config file and run Weston.

    # Start a new ssh session
    mkdir -p /etc/xdg/weston
    touch /etc/xdg/weston
    vi /etc/xdg/weston/weston.ini

    Paste the following options into weston.ini to allow Weston to run without an output display.

    # configuration file for Weston
    [core]
    idle-time=0
    backend=drm-backend.so
    repaint-window=10
    gbm-format=xbgr8888
    require-outputs=none
    

    [output]
    name=DSI-1
    mode=on
    

    [output]
    name=DP-1
    mode=on
    

    [shell]
    clock-format=seconds

    Now run Weston as indicated in the Qualcomm Linux documentation:

    export GBM_BACKEND=msm && export XDG_RUNTIME_DIR=/dev/socket/weston && mkdir -p $XDG_RUNTIME_DIR && weston --continue-without-input --idle-time=0

  9. Start the GStreamer AI multi input output object detection sample application with the instructions from the Qualcomm Linux AI sample application documentation.

    # Start a new ssh session
    # Setup some environment variables in the SSH session for display
    export XDG_RUNTIME_DIR=/dev/socket/weston && export WAYLAND_DISPLAY=wayland-1
    

    # Run the GStreamer AI multi input output object detection sample application
    gst-ai-multi-input-output-object-detection --num-rtsp=1 --rtsp-ip-port=127.0.0.1:8901/mystream --out-rtsp -i 127.0.0.1 -p 8900

  10. Playback the output stream on local machine with FFplay (part of FFmpeg):

    ffplay -probesize 32 -analyzeduration 0 -sync ext -fflags nobuffer -flags low_delay -x 640 -y 480 -rtsp_transport tcp rtsp://127.0.0.1:8900/live

    Alternatively, use a video player like VLC that can play RTSP streams from over the network.

Previous
Next
Terms of UsePrivacy PolicyCookie PolicyCookie SettingsContact Us
2025 ©Qualcomm Technologies, Inc. and/or its affiliated companies. All rights reserved. References to "Qualcomm" may mean Qualcomm Incorporated, or subsidiaries or business units within the Qualcomm corporate structure, as applicable. Qualcomm Incorporated includes Qualcomm's licensing business, QTL, and the vast majority of its patent portfolio. Qualcomm Technologies, Inc., a wholly-owned subsidiary of Qualcomm Incorporated, operates, along with its subsidiaries, substantially all of Qualcomm's engineering, research and development functions, and substantially all of its products and services businesses. Qualcomm products referenced on this page are products of Qualcomm Technologies, Inc. and/or its subsidiaries. Materials that are as of a specific date, including but not limited to press releases, presentations, blog posts and webcasts, may have been superseded by subsequent events or disclosures. Nothing in these materials is an offer to sell any of the components or devices referenced herein.