Curvy facades

[vvoovv]

Classification of curvy facades produces odd results:

(bern_old_town.osm, Kirche St. Peter und Paul):

[vvoovv]

(bern_old_town.osm, Kirche St. Peter und Paul):

[vvoovv]

I suggest skipping curvy facades from from the visibility and classification calculations.

[vvoovv]

How to define a curvy facade?

A curvy facade is a sequence of narrow facades connecting each other "smoothly".

[polarkernel]

A curvy facade is a sequence of narrow facades connecting each other "smoothly".

Are we back to apse detection (see prochitecture/bpypolyskel#5 (comment))?

[vvoovv]

A curvy facade is a sequence of narrow facades connecting each other "smoothly".

Are we back to apse detection (see prochitecture/bpypolyskel#5 (comment))?

A curvy facade can have an arbitrary shape.

[vvoovv]

I suggest skipping curvy facades from from the visibility and classification calculations.

That was the wrong idea.

A front part must be still detected to place an entrance. For example, suppose we have a building with a circular footprint (it could be an ancient tower). It still needs an entrance.

[vvoovv]

Curvy sequences can be also concave. Example.

[vvoovv]

A curvy sequence must be classified if

1. there is no entrance tagged in OSM for all edges
   AND
2. there are no wide edges classified as the front facades

One approach would be to connect the first and the last node of the curve sequence with a straight line and use that proxy edge for the visibility calculation and classification. But how would it work for a concave sequence if a way follows the concave sequence and is located very close to the concave sequence?

[vvoovv]

A special case is a completely curvy building. Example: a circular footprint.

[vvoovv]

Are we back to apse detection (see prochitecture/bpypolyskel#5 (comment))?

The apse detection and the detection of other patterns will required to generate realistic models of churches and cathedrals.

[vvoovv]

Many stadiums have curvy footprint.

[vvoovv]

Joined with #16.