This folder contains the output of filtering and node tracking based on the raw graphs obtained by NEFI. The raw graphs were first filtered and then subjected to node tracking. The tracked graphs are identical to the filtered graphs but receive an additional node property, namely color. Filtering reduces the size of the LP and increases the distance between neighboring nodes which is beneficial for the accuracy of the tracking. In particular we have used two filter strategies based on the high data sets:

high
high_unfiltered

high:

All edges that are not part of a cycle are removed. All edges that are shorter than 30 pixel are contracted. All nodes of degree 2 are smoothed. Only the largest connected component is kept. 

This filter leads to graphs that have excellent properties for node tracking whilst capturing the overall topology of physarum well. They can be filtered further to focus on edges of certain properties.

high_unfiltered:

Despite the (unhappy) name, filters have been used on these graphs. All edges that are not part of a cycle are removed. Edges of width smaller than 5 pixels are removed. Only the largest connected component is kept.

These filter settings lead to graphs that trace the face structure of physarum extremely well due to long chains of degree 2 nodes being present. We remove edges that are very thin, because we do not trust the resolution of NEFI on this level. By doing so we introduce a small error in topology, but likely achieve larger accuracy on the edge properties. 

These examples for filtering illustrate the need to tailor filters to the relevant question one wants to investigate. Also, this hints at the large amount of information that is still waiting to be extracted from this dataset.