Mtriage development has been stalled for some time, as I've had other priorities. I'm hoping to pick up the pace in the last couple of months of 2020, however. These are a couple of things on the near horizon:
This is pretty key to ensure that community component contributions are meaningful. Currently there are only tests for the core part of mtriage, and it is just an article of faith that each component (selector/analyser) works.
None of our code for bootstrapping custom classifiers with mtriage is public, which means that it's pretty much useless to everyone in its current state, except as a reference. We're looking to drive some more research around training vision classifiers using synthetic data with mtriage, and we'll fold out all of these fixes into upstream mtriage as ways to apply custom classifiers in the abstract via mtriage as a deployment framework.
This is somewhat irresponsible on my part, as this here Python/Docker version of the framework barely works. But I'm looking for a way to get my fingers properly sticky with Rust in practice, and mtriage seems a good candidate (systems software, containers, etc).