eduzhai > Applied Sciences > Engineering >

Scalable Proposal-free Instance Segmentation Network for 3D Pixel Clustering and Particle Trajectory Reconstruction in Liquid Argon Time Projection Chambers

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 16 pages

Abstract: Liquid Argon Time Projection Chambers (LArTPCs) are high resolution particleimaging detectors, employed by accelerator-based neutrino oscillationexperiments for high precision physics measurements. While images of particletrajectories are intuitive to analyze for physicists, the development of a highquality, automated data reconstruction chain remains challenging. One of themost critical reconstruction steps is particle clustering: the task of grouping3D image pixels into different particle instances that share the same particletype. In this paper, we propose the first scalable deep learning algorithm forparticle clustering in LArTPC data using sparse convolutional neural networks(SCNN). Building on previous works on SCNNs and proposal free instancesegmentation, we build an end-to-end trainable instance segmentation networkthat learns an embedding of the image pixels to perform point cloud clusteringin a transformed space. We benchmark the performance of our algorithm onPILArNet, a public 3D particle imaging dataset, with respect to commonclustering evaluation metrics. 3D pixels were successfully clustered intoindividual particle trajectories with 90 of them having an adjusted Rand indexscore greater than 92 with a mean pixel clustering efficiency and purity above96 . This work contributes to the development of an end-to-end optimizable fulldata reconstruction chain for LArTPCs, in particular pixel-based 3D imagingdetectors including the near detector of the Deep Underground NeutrinoExperiment. Our algorithm is made available in the open access repository, andwe share our Singularity software container, which can be used to reproduce ourwork on the dataset.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×