docker container with opencv + gstreamer in python for openHD

This is a tutorial how to setup a docker container containing opencv with gstreamer integration for usage in python e.g. for openHD. Tested with windows host.

build container

docker build -t opencv-test .

run container

• with windows file system folder mounted

winpty docker run -v //c/data:/home/data -p 5000:5000/udp -p 5004:5004/udp -it opencv-test

• with wsl file system folder mounted
  • this is faster since docker and wsl share the same filesystem

winpty docker run -v //wsl$/Ubuntu-20.04/home/<user_name>/data:/home/data -it opencv-test

• note: winpty is required when executing on windows from git-bash
• port 5000 is the videao stream
• port 5004 is mavlink

setup wsl

in power shell:

wsl --set-default-version 2
wsl --install -d Ubuntu-20.04
wsl -l -o // options
wsl -l -v // installed

// to delete it:
wsl --unregister Ubuntu-20.04

in explorer: \\wsl$

testing

• check for data traffic : netcat -ul 5000
• send test signal from console:

// sample signal
gst-launch-1.0 videotestsrc ! x264enc ! video/x-h264, stream-format=byte-stream ! udpsink host=127.0.0.1 port=5000
// from cam (untested)
gst-launch-1.0 v4l2src device=/dev/video0 ! h264parse ! avdec_h264 ! udpsink host=127.0.0.1 port=5000

• receive signal from console source:

gst-launch-1.0 udpsrc port=5000 ! h264parse ! avdec_h264 ! autovideosink sync=false

check for udp traffic:

apt-get install tcpdump
tcpdump -n udp port 5000

run opencv + gstreamer in python

• create a new file test.py in your mounted data folder (e.g. C:/data)

import time
import cv2

cap = cv2.VideoCapture("udpsrc port=5000 ! h264parse ! decodebin ! videoconvert ! appsink", cv2.CAP_GSTREAMER)
while True:
    ret,frame = cap.read()
    if not ret:
        print("no frame")
        time.sleep(1)
        continue
    print("got frame")
    cv2.imwrite('test.jpg', frame)
cv2.destroyAllWindows()
cap.release()

• connect your wifi to the openHD network, pw: wifiopenhd
• run the container and the python file

winpty docker run -v //c/data:/home/data -p 5000:5000/udp -it opencv-test
cd /home/data
python3 test.py

• end python script by pressing ctrl + C