I wanted an overlay with my heart rate for a Zoom fitness class I take, because why not? So I looked into using OBS with v4l2loopback. It turns out that writing a custom video input source for OBS is a huge pain and can't be done in Python, and although I love Lua, it's not really batteries-included enough to do what I want (I want to use GATT in some way to pull from a HRM and Cairo to draw on an image, and that seems hard to come up with a reproducible environment for in Lua).
It turns out that this is an enormous can of worms and literally no single
part of this is easy. Cairo in Python is a huge pain in the butt and
requires a lot of format conversions (RGB24, well, isn't). There appears
to actually be no good way to talk to a GATT device from Python on Linux.
Or at all on Linux, for that matter. I tried using pygatt, and it
inexplicably cannot reliably connect to a BLE device sometimes without first
being helped along by gatttool, so instead I use someone else's script
that does this huge ball of wax with pyexpect to talk to gatttool
directly (!!). And pyfakewebcam requires RGB input -- which it then
immediately converts to YUV -- so I have to do two format conversions (one
from Cairo's bizarre BGRA format to RGB, one from RGB to YUV), which is
needlessly slow.
Anyway this took a lot longer than I had hoped and I got to reuse a lot less than I hoped. I was really hoping to just glue things together. That extremely did not happen. Oh well. Sometimes you get the elevator and sometimes you get the shaft.
To set this up:
sudo apt-get install python3-gst-1.0pip3 install -r requirements.txtsudo modprobe v4l2loopback devices=1 video_nr=20 card_label="v4l2loopback" exclusive_caps=1sudo hcitool lescanto find the BLE address of your HRM- Edit camcairo.py with your parameters (should I have an argument parser? probably!)
python camcairo.py- To test:
ffplay /dev/video20
Based on https://github.com/fg1/BLEHeartRateLogger . Implementation ideas from https://elder.dev/posts/open-source-virtual-background/ , though I ended up using Gstreamer because I initially blamed OpenCV for only getting 15fps out of my camera, but in the light of day discovered that I am a huge idiot and the reason I was only getting 15fps was because there was not enough light for the camera to do 30fps.