Now that we have a much smaller dataset for testing purposes, we can probably include training in the CI tests. We only have <100 images (maybe we can even further reduce the size), it should be fairly quick to complete a few epochs, then compare model accuracies.
We can then try switch between those newly created models to double check their performances.
Part of Issue https://github.com/PaulCCCCCCH/Robustar_implementation/issues/186
Now that we have a much smaller dataset for testing purposes, we can probably include training in the CI tests. We only have <100 images (maybe we can even further reduce the size), it should be fairly quick to complete a few epochs, then compare model accuracies.
We can then try switch between those newly created models to double check their performances.