Open tibuch opened 7 years ago
@ctrueden and @dietzc why do we have expensive geom tests and normal ones?
I think for historical reasons. If you find "normal" ones which cover all the cases, it's absolutely OK.
@tibuch I thought you were the one who authored those expensive tests! No?
I would much rather have only fast-running tests but which still exercise the algorithms sufficiently. But my understanding was that someone (you? someone else?) thought the small ones might not be rich/complex enough to fully exercise the algorithms?
@ctrueden I guess the tests were added by @seebacherd.
I will replace the images by smaller but complexer (curves, holes and so on) ones.
I would say the accuracy of the tests depends a lot more on small epsilons.
@tibuch Thanks, sounds great!
Test and improve FeatureSets.