Robert Sachunsky

> In the light of the statement cited above and to cover not only Python 3.8 environments but also non-Intel hosts, I think we need separate checks for each sub-env....

> The test is much simpler: just try to install one of the required Python packages (keras, tensorboard, tensorflow, ...). You mean in the Makefile, bypassing the module's `setup.py` or...

> A sub-env for TensorFlow 1 like the current `headless-tf1` can try to install a Python module for TensorFlow 1 without anything going more against encapsulation than we already have...

> > our recipes would look like this > > No, that would still try installing TF1 several times. No it would not. Remember, we use `multirule` above, so this...

> I just read the Dependabot alerts for `cor-asv-ann`. It lists 218 (!) alerts for `tensorflow-gpu` with 3 of them marked critical and 60 of them marked high. That's a...

Agreed! So in the end we'll just have to start trying – and back every step of the way up by _some_ tests at least. BTW, for 3.8 we have...

BTW, `shapely.geometry.polygon.Polygon` has very nice API for the first 2 tasks, including `contains()` and `area()`. The third could be achieved with ad-hoc binarization and some simple Numpy statistics like `count_nonzero()`...

A good reference for additional checks are the validation error classes in [Aletheia](https://www.primaresearch.org/www/assets/tools/Aletheia%20User%20Guide.pdf), p. 118/119.

> c.f. https://github.com/OCR-D/ocrd_evaluate_segmentation now renamed to https://github.com/OCR-D/ocrd_segment (there will be more processors)

Moved the original issue from core here to have a better reminder of what is left to do. Out of the original list, we are still somewhere in the first...