You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Reduce irregular boundaries in final roof partition
Reduce complexity of initial roof partition
Todo:
Line regularisation: cluster on distance in general between line segments, not just perpendicular distance
Line regularisation: cluster with moving average for each cluster/implement proper hierarchical clustering
compute more accurate extend for intersection lines
In initial line roof partition (the arrangement prior to graph cut optimisation): insert only the line segments, not full lines
do not merge clusters if both contain an intersection line (or split again after clustering?)
test on large area
remove complexity limit in graph cut optimisation node. Maybe put it in the regularisation node...
Make new clustering algorithm faster. 1) Use a binary min heap for the distance table/finding minimum distances, 2) keep for each cluster a std::list with heap handles to the distances involving the cluster (for faster distance update after a merge).
The text was updated successfully, but these errors were encountered:
Goals:
Todo:
Line regularisation: cluster on distance in general between line segments, not just perpendicular distance
Line regularisation: cluster with moving average for each cluster/implement proper hierarchical clustering
compute more accurate extend for intersection lines
In initial line roof partition (the arrangement prior to graph cut optimisation): insert only the line segments, not full lines
do not merge clusters if both contain an intersection line (or split again after clustering?)
test on large area
remove complexity limit in graph cut optimisation node. Maybe put it in the regularisation node...
Make new clustering algorithm faster. 1) Use a binary min heap for the distance table/finding minimum distances, 2) keep for each cluster a std::list with heap handles to the distances involving the cluster (for faster distance update after a merge).
The text was updated successfully, but these errors were encountered: