Working on a UI test automation tool that leverages vision-based UI-perception and semantics, runs purely locally on CPU. It's called VizQA: https://github.com/TinyReasonLabs/vizQA

This solves both the laziness of creating UI test automation setups, without it breaking everytime a slight change is made. with VizQA, the test is just a yaml file describing sequences of simple steps, to navigate, interact, and make assertions.

I just made an initial release, looking for feedback and opinions for such a powerful tool, also open for contribution!