Cargando…

Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data

High-energy physics detectors, images, and point clouds share many similarities in terms of object detection. However, while detecting an unknown number of objects in an image is well established in computer vision, even machine learning assisted object reconstruction algorithms in particle physics...

Descripción completa

Detalles Bibliográficos
Autor principal: Kieseler, Jan
Lenguaje:eng
Publicado: 2020
Materias:
Acceso en línea:https://dx.doi.org/10.1140/epjc/s10052-020-08461-2
http://cds.cern.ch/record/2711960
_version_ 1780965264763912192
author Kieseler, Jan
author_facet Kieseler, Jan
author_sort Kieseler, Jan
collection CERN
description High-energy physics detectors, images, and point clouds share many similarities in terms of object detection. However, while detecting an unknown number of objects in an image is well established in computer vision, even machine learning assisted object reconstruction algorithms in particle physics almost exclusively predict properties on an object-by-object basis. Traditional approaches from computer vision either impose implicit constraints on the object size or density and are not well suited for sparse detector data or rely on objects being dense and solid. The object condensation method proposed here is independent of assumptions on object size, sorting or object density, and further generalises to non-image-like data structures, such as graphs and point clouds, which are more suitable to represent detector signals. The pixels or vertices themselves serve as representations of the entire object, and a combination of learnable local clustering in a latent space and confidence assignment allows one to collect condensates of the predicted object properties with a simple algorithm. As proof of concept, the object condensation method is applied to a simple object classification problem in images and used to reconstruct multiple particles from detector signals. The latter results are also compared to a classic particle flow approach.
id cern-2711960
institution Organización Europea para la Investigación Nuclear
language eng
publishDate 2020
record_format invenio
spelling cern-27119602023-09-27T07:55:50Zdoi:10.1140/epjc/s10052-020-08461-2http://cds.cern.ch/record/2711960engKieseler, JanObject condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image datahep-exParticle Physics - Experimenteess.IVcs.CVComputing and Computersphysics.data-anOther Fields of PhysicsHigh-energy physics detectors, images, and point clouds share many similarities in terms of object detection. However, while detecting an unknown number of objects in an image is well established in computer vision, even machine learning assisted object reconstruction algorithms in particle physics almost exclusively predict properties on an object-by-object basis. Traditional approaches from computer vision either impose implicit constraints on the object size or density and are not well suited for sparse detector data or rely on objects being dense and solid. The object condensation method proposed here is independent of assumptions on object size, sorting or object density, and further generalises to non-image-like data structures, such as graphs and point clouds, which are more suitable to represent detector signals. The pixels or vertices themselves serve as representations of the entire object, and a combination of learnable local clustering in a latent space and confidence assignment allows one to collect condensates of the predicted object properties with a simple algorithm. As proof of concept, the object condensation method is applied to a simple object classification problem in images and used to reconstruct multiple particles from detector signals. The latter results are also compared to a classic particle flow approach.High-energy physics detectors, images, and point clouds share many similarities in terms of object detection. However, while detecting an unknown number of objects in an image is well established in computer vision, even machine learning assisted object reconstruction algorithms in particle physics almost exclusively predict properties on an object-by-object basis. Traditional approaches from computer vision either impose implicit constraints on the object size or density and are not well suited for sparse detector data or rely on objects being dense and solid. The object condensation method proposed here is independent of assumptions on object size, sorting or object density, and further generalises to non-image-like data structures, such as graphs and point clouds, which are more suitable to represent detector signals. The pixels or vertices themselves serve as representations of the entire object, and a combination of learnable local clustering in a latent space and confidence assignment allows one to collect condensates of the predicted object properties with a simple algorithm. As proof of concept, the object condensation method is applied to a simple object classification problem in images and used to reconstruct multiple particles from detector signals. The latter results are also compared to a classic particle flow approach.arXiv:2002.03605oai:cds.cern.ch:27119602020-02-10
spellingShingle hep-ex
Particle Physics - Experiment
eess.IV
cs.CV
Computing and Computers
physics.data-an
Other Fields of Physics
Kieseler, Jan
Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
title Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
title_full Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
title_fullStr Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
title_full_unstemmed Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
title_short Object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
title_sort object condensation: one-stage grid-free multi-object reconstruction in physics detectors, graph and image data
topic hep-ex
Particle Physics - Experiment
eess.IV
cs.CV
Computing and Computers
physics.data-an
Other Fields of Physics
url https://dx.doi.org/10.1140/epjc/s10052-020-08461-2
http://cds.cern.ch/record/2711960
work_keys_str_mv AT kieselerjan objectcondensationonestagegridfreemultiobjectreconstructioninphysicsdetectorsgraphandimagedata