When doing the final combining of clusters across neighborhoods (i.e. after phase 2), it would be useful to record the source neighborhood for each cluster. This value should generally be the same as the channel on which the template has its max value, but we are trying to debug a case of cluster 'splitting', and we think that in some circumstances these may diverge.
When doing the final combining of clusters across neighborhoods (i.e. after phase 2), it would be useful to record the source neighborhood for each cluster. This value should generally be the same as the channel on which the template has its max value, but we are trying to debug a case of cluster 'splitting', and we think that in some circumstances these may diverge.
This value can't be recomputed later via metrics.