Client side web based opencv face detection
With this project I wanted to see how much of opencv can be ported to work
on emscripten with .wasm.
Live demo here
There are two main parts to in the codebase /web and /cpp.
index.html in /web loads a video in a <video> tag and runs a small script
that copies each frame of that video onto a canvas, runs getImageData on that canvas
and passes the resulted pixel array to c++.
main.cpp in /cpp grabs the pixel array passed by js, wraps it in an opencv Mat
and runs haar face detection on it.
Afterwards the image is displayed using emscripten’s port of SDL.
To build this project you will need emscripten which can be found here.
You will also need a version of Opencv built with emscripten, you can either download the version I build it here,
or build it yourself (instructions below).
After getting the dependencies you will need to set 2 environment variables
EMSCRIPTEN=emscripten directory, if you run source ./emsdk_env.sh this will be set automatically
After this just go ahead and
cd cmake-debug-build # or another directory if you wantcamke ..make
This will build the project, and make an index.html, it needs a webserver to work properly.
You could use python
python -m SimpleHTTPServer
Open localhost:8000 in your browser and thats it!
Coming soon…
use prebuilt .zip for now