backend: Resolve incompatibilities between the Python backend and the Raspberry Pi 5

[ethanjli]

Motivation

Currently, the Python hardware controller is unable to run on the Raspberry Pi 5 - it depends on the rpi-gpio package, which does not work on the Raspberry Pi 5 (see https://adafruit-playground.com/u/MakerMelissa/pages/comparing-libgpiod-and-gpiozero-speeds-on-the-raspberry-pi-5 for details). Additionally, the Python segmenter depends on scikit-image and opencv-contrib-python-headless, which do not have any available piwheel (i.e. armv7l) builds on bookworm.

Goals

• Make the Python hardware controller able to run on the Raspberry Pi 5, ideally with the 64-bit version of the Raspberry Pi OS
• Make the Python segmenter able to run on the Raspberry Pi 5, ideally with the 64-bit version of the Raspberry Pi OS
• If possible, maintain compatibility with the last 32-bit version of Raspberry Pi OS 11 (bullseye), which was the base for PlanktoScope software v2023.9.0

Steps

• os/pallet: Containerize the Python segmenter and deliver it via Forklift #326
• Migrate from the rpi-gpio package to some other GPIO library

Unresolved Questions

Implementation History

• Deploy the backend segmenter with forklift rather than the distro setup scripts #352
• Migrate to the picamera2 library for camera control & image acquisition device-backend#19

[ethanjli]

It might be possible to containerize the Python hardware controller using https://github.com/dtcooper/raspberrypi-os-docker as a base image, but I'm not sure whether this would specifically solve the rpi-gpio problem since the difference between the Pi 4 and the Pi 5 is a different memory mapping for GPIO.

[ethanjli]

RPi5-compatible alternatives to rpi-gpio include:

• gpiod, which has examples in-the-wild of usage in a Docker container: https://github.com/devdotnetorg/docker-libgpiod
• gpiozero, which has rudimentary examples in-the-wild of usage in a Docker container: https://forums.balena.io/t/getting-gpio-to-work-in-container/367928/5
• rpi-lgpio, which probably would not require extensive changes to the hardware controller, but which is not widely used and does not have any examples in-the-wild of usage in a Docker container

[ethanjli]

@Oumayma-hy As we rewrite the Python hardware controller, let's try to do it in a way that reduces the amount of additional work needed to make it compatible with the Raspberry Pi 5, i.e. by only using Python packages which other people have found to work on the Raspberry Pi 5. Probably 80% of this is just a matter of switching from the rpi-gpio library to the gpiod library (if that library is fine) or else the gpiozero library (if gpiod really does not meet our needs).