Alex Vorndran

Before working on the issue there should be consens which of the two variants is supposed to be used in the implementation and as a consequence be documented. @SergioRAgostinho What's...

@taketwo Yeah it seems like the class is inconsistent since that commit in 2011. To be honest, I'm not sure how the model is different from the cylinder model. I...

### Broken uses of the model coefficients Broken in the sense that their use of the model coefficients contradicts `computeModelCoefficients`. - `getDistancesToModel` (used e.g in LMEDS, MLESAC, MSAC, RMSAC) -...

@taketwo I think the main difference is supposed to be that the stick model can work without normals. But the implementation also does not seem to compute a radius/width value...

I had time to fix the model coefficients in my fork so that they work (mostly) consistent to the documentation and all the other similar models again. At the moment...

As a follow up to previous comment I would like to add that there are indeed some subtle differences between the stick and the line model. In my application scenario...

@mvieth I follow most of your observations, but would like to mention that in the cylinder model the radius is also determined based on only two sample points (and some...

> Ah, interesting. My guess would be that the line model with a bit of threshold-parameter-tuning would work for that scenario, no? Yes, you are right. This is how I...

Someone gave it a try and put a version up on conda-forge: https://github.com/conda-forge/commitizen-feedstock/. Interestingly (or sadly?), it was last updated one day before this issue was opened.