GammaBoard is a dashboard based on jupyter notebooks technology developed to display specific metrics assessing the reconstructions performances of Imaging Atmospheric Cherenkov Telescopes (IACTs).

Deep learning research is very experimental and is a lot about trials and errors, and bookkeeping of the different experiments realised. GammaBoard eases the bookkeeping and allows quick comparison of the reconstruction performances of your machine learning experiments.
It is a working prototype used in CTA, especially for the GammaLearn project.

GammaBoard is composed of :

  • Plots (metrics) such as angular resolution and energy resolution.
  • The list of experiments in the user folder.

When an experiment is selected in the list, the data is automatically loaded, the metrics computed and displayed. A list of information provided during the training phase is also displayed.
As many experiment results as needed can be overlaid.


Thomas Vuillaume and Mikael Jacquemont