Duplication

After the combined refinement, PartFinder removes duplicates of detections by applying an additional clustering to the results of the combined refinement.

The duplication removal is similar to the clustering of hypotheses (see documentation for the Cluster node) with the following exceptions:

• the duplication removal utilizes hierarchical clustering with respect to the CombinedScore,

• there are no limits for the number of clusters or a cluster’s leading score,

• the size of a cluster is one,

• the scale of a cluster’s spatial vicinity area can be set via Scaling parameter.

Format

Object

Please refer to the sub node documentation.