NUK - logo
E-viri
Recenzirano Odprti dostop
  • Semantic segmentation with ...
    Abratenko, P.; Asaadi, J.; Ashkenazi, A.; Balasubramanian, S.; Baller, B.; Barnes, C.; Barr, G.; Bhanderi, A.; Bishai, M.; Caro Terrazas, I.; Castillo Fernandez, R.; Chen, Y.; Church, E.; Cianci, D.; Conrad, J. M.; Convery, M.; Cooper-Troendle, L.; Crespo-Anadón, J. I.; Del Tutto, M.; Dennis, S. R.; Devitt, A.; Diurba, R.; Eberly, B.; Ereditato, A.; Fleming, B. T.; Franco, D.; Furmanski, A. P.; Gollapinni, S.; Greenlee, H.; Gu, W.; Guenette, R.; Guzowski, P.; Hagaman, L.; Hall, E.; Hen, O.; Horton-Smith, G. A.; Hourlier, A.; James, C.; Ji, X.; Jo, J. H.; Johnson, R. A.; Kaneshige, N.; Karagiorgi, G.; Ketchum, W.; Kirby, B.; Kreslo, I.; LaZur, R.; Li, K.; Li, Y.; Littlejohn, B. R.; Louis, W. C.; Marchionni, A.; Mariani, C.; Marsden, D.; Martin-Albo, J.; Mastbaum, A.; Miller, K.; Mills, J.; Mogan, A.; Mooney, M.; Moore, C. D.; Mora Lepin, L.; Mousseau, J.; Naples, D.; Palamara, O.; Papavassiliou, V.; Pate, S. F.; Paudel, A.; Pavlovic, Z.; Ponce-Pinto, I. D.; Raaf, J. L.; Radeka, V.; Rafique, A.; Rochester, L.; Rogers, H. E.; Rosenberg, M.; Russell, B.; Scanavini, G.; Schmitz, D. W.; Schukraft, A.; Seligman, W.; Sharankova, R.; Smith, A.; Soleti, S. R.; Spitz, J.; John, J. St; Sword-Fehlberg, S.; Tang, W.; Thorpe, C.; Toups, M.; Tsai, Y.-T.; Uchida, M. A.; Usher, T.; Van De Pontseele, W.; Viren, B.; Wolbers, S.; Yang, T.; Yates, L. E.; Zeller, G. P.; Zennamo, J.

    Physical review. D, 03/2021, Letnik: 103, Številka: 5
    Journal Article

    We present the performance of a semantic segmentation network, sparsessnet, that provides pixel-level classification of MicroBooNE data. The MicroBooNE experiment employs a liquid argon time projection chamber for the study of neutrino properties and interactions. sparsessnet is a submanifold sparse convolutional neural network, which provides the initial machine learning based algorithm utilized in one of MicroBooNEs νe-appearance oscillation analyses. The network is trained to categorize pixels into five classes, which are reclassified into two classes more relevant to the current analysis. The output of sparsessnet is a key input in further analysis steps. This technique, used for the first time in liquid argon time projection chambers data and is an improvement compared to a previously used convolutional neural network, both in accuracy and computing resource utilization. The accuracy achieved on the test sample is ≥ 99 %. For full neutrino interaction simulations, the time for processing one image is ≈ 0.5 sec , the memory usage is at 1 GB level, which allows utilization of most typical CPU worker machine.