Slicing with deep learning models at ProtoDUNE-SP

Abstract DUNE is a cutting-edge experiment aiming to study neutrinos in detail, with special focus on the flavor oscillation mechanism. The prototype of Far Detector Single Phase TPC (ProtoDUNE-SP) was built and operated at CERN full set reconstruction tools. To implement these tools, Pandora, multi-algorithm framework, has been developed. A large number algorithms, some them being exploiting traditional clustering, detector physics deep learning approaches, have applied images gradually build up picture out singular events. One such algorithms Pandora slicing algorithm which aims partition hits an event sets called slices. Each slice represents single interaction should identify all related interacting particle its subsequent decay products. We expect order tens slices per ProtoDUNE-SP. In this paper we present approach problem, designing model able outperform state-of-the-art currently implemented within Pandora. assess performance our tool terms efficiency accuracy, while hardware accelerating setups. ultimate goal incorporate tool.