Two places in Pandora reconstruction where this is useful: characterisation of 2D Clusters before shower growing algorithm attempts to add branches to shower seeds (provide a hint to later algorithms) and at end of reconstruction, classifying final Particles based on their constituent Clusters.
#2 Updated by John Marshall about 3 years ago
- Assignee changed from Lorena Escudero sanchez to John Marshall
- % Done changed from 30 to 60
Work performed by John and Lorena.
Provide track vs. shower identification metrics at three key points in the reconstruction:
1. Characterise Particles made by initial 3D track reconstruction pass; 3x2D Clusters available for examination, no "branches" will have been added to Clusters at this stage. Remove Shower Particles at this stage.
2. Characterise 2D Clusters not present in a Particle, prior to shower branch growing algorithm. Individual 2D Clusters available for examination, no "branches" will have been added at this stage.
3. Characterise final Particles to provided output track vs. shower identification; 3x2D Clusters available for examination, post shower branch addition.
Many topological variables calculated and written to ROOT TTrees at each step. Output trees analysed by TMVA, with view to using only a selection of optimised rectangular cuts. Significant gains made (as determined by Pandora pattern recognition metrics) for MicroBooNE events, especially for CCRES interactions w/ pizeros.
It is expected that these algorithms will need to be re-optimised for different use-cases. Nick Grant (Warwick) will soon begin to look into this for the DUNE use-cases.
Not yet released to larsoft; expect to do this within next couple of weeks.