Gamma-ray trajectory reconstruction using Deep Learning methods

The Dark Matter Particle Explorer experiment allows for $\gamma$-ray detection up to TeV energies, with an unprecedented energy resolution of about 1\%, which makes it a unique instrument for $\gamma$-ray physics at these energies. Diffuse $\gamma$-ray analysis requires a precise tracking method in order to identify the incoming particle through the different detectors, as well as defining its origin in space. A deep-learning tool for track reconstruction has already been developed for electrons and ions. In this analysis we used this tool trained on electron samples on $\gamma$-ray samples to assess its efficiency on trajectory reconstruction up to 10~TeV. Preliminary results show very promising tracking precision and bring the prospect of this new tool in high energy $\gamma$-ray study. The efficiency of the $\gamma$ selection by the deep-learning method is compared to the original track of simulated events.