Add logic to validate the outcome of the examples #37
Description
Currently, we're checking that the examples pass and that the images and segmentation look appropriate visually. We'll likely still change some parameters over time, thus I wouldn't expect to fully reproduce the same results (e.g. the segmentation parameters aren't optimized for all examples yet => makes sense to change them).
It would nevertheless make sense to have a script we can run for some examples (e.g. just example 01 to start with) which checks some of the content of the results.
Ideas for what to check:
- Is the OME-Zarr structure still the same (how to we serialize the structure?)
- Do we get the same number of segmented objects (unique labels or number of measurements in the measurement table)
- Either compare full measurement tables or some summary statistics (for 01 feasible to just check against the expected table. Wouldn't make sense for larger examples)
Those then are a bit like integration tests, but we probably don't want to run them at every commit. Can be valuable though to run e.g. before some new releases.