PILArNet

PILArNet是由深度学习物理合作组织创建的，旨在支持高能物理实验中液氩时间投影室（LArTPC）数据分析的公共数据集。该数据集包含300,000个模拟样本，涵盖三种不同体积大小的数据，采用稀疏2D和3D矩阵格式存储，并附带模拟粒子的辅助信息。数据集的创建过程涉及使用Geant4和LArSoft进行粒子跟踪模拟，并通过空间模糊处理模拟真实探测器分辨率。PILArNet的应用领域主要集中在粒子图像重建和分析技术的发展，旨在解决高分辨率粒子轨迹图像中提取高级物理信息的技术挑战。PILArNet is a public dataset developed by the Deep Learning Physics Collaboration to support data analysis for Liquid Argon Time Projection Chamber (LArTPC) experiments in high-energy physics. This dataset comprises 300,000 simulated samples, covering data across three distinct volume sizes, stored in sparse 2D and 3D matrix formats, and includes supplementary information for simulated particles. The dataset was developed using Geant4 and LArSoft for particle tracking simulations, with spatial blurring applied to emulate the resolution characteristics of real detectors. The primary applications of PILArNet focus on advancing particle image reconstruction and analysis technologies, aiming to resolve technical challenges associated with extracting high-level physical information from high-resolution particle trajectory images.

搜集汇总

背景与挑战

背景概述

PILArNet是一个由DeepLearnPhysics维护的公开数据集，专注于液态氩时间投影室（LArTPC）的模拟3D数据，旨在支持粒子物理中的像素级分类技术开发。该数据集适用于计算机视觉任务，如3D聚类、实例分割和语义分割，采用CC-By Attribution 4.0 International许可证共享。