A framework in C# to evaluate road graphs inferred by trajectory-based road map construction algorithms. Part of master thesis for the University Utrecht Department of Information and Computing Science. This version works without visualisations. We heavily recommend the visual version, which you can find here.
- Import directed, undirected and hybrid maps.
- Transform directedness and size of imported maps.
- Export edited maps.
- Import trajectory sets.
- Synthetically generate shortest path trajectories for imported maps.
- Export generated trajectory sets.
- Evaluate similarity between two imported maps using the graph sampling similarity measure by Biagioni and Eriksson.
We follow a common format in the map construction field, similar to the one used in data-sets on Mapconstruction. The following format is expected for maps.
A Vertices file, with on each line:
vertexId, projectedX, projectedY
An Edges file, with on each line:
edgeId, vertexId1, vertexId2, directedness
Trajectory sets are expected in a folder containing a trip_x file for each trajectory, with on each line:
projectedX projectedY time
We provide Projectors to help with conversion of coordinates. Additionally, with help of external tools, extracted Open Street Map road graphs can be used. The OSMParser helps with the final small step to get such a graph to our format.
The example below shows a sketch of how a map can be evaluated.
Map gt = importMap.ReadMap($"data/Ground-truth");
Map cm = importMap.ReadMap($"data/Constructed-map");
//Evaluates the map by averaging 100 local neighbourhoods.
(float, float) similarityScore = eval.EvalMap(gt, cm, 100);
Console.WriteLine(similarityScore);