AttributeClusterization - Extended-Object-Detection-ROS/wiki_english GitHub Wiki

Clustering

For some attributes, the exit may require additional manipulations, incl. clustering areas. For this, the package provides the appropriate functionality. Note that clustering will be applied to the output of the attribute BEFORE filters, if any.

You can add clustering to an attribute by writing a description of the method in its tag (the one in AttributeLib).

<Attribute Name="HistSkin" Type="HistColor" Histogram="histograms/Skin.yaml">    
        <Clusterization Type="FOREL" R="100" eps="10"/>
</Attribute>

The Clusterization tag contains the Type parameter, which is responsible for the clustering method. Currently available:

  • FOREL () An algorithm for clustering objects in places of their greatest concentration. Has parameters:
    • R (double, default: 0) Cluster radius.
    • eps (double, default: 0) Acceptable convergence threshold.
⚠️ **GitHub.com Fallback** ⚠️