SPARKESX: Single-dish PARKES data sets for finding the uneXpected - Part 2

We present the Single-dish PARKES data sets for finding the uneXpected (SPARKESX), a compilation of real and simulated high-time resolution observations. SPARKESX comprises three mock surveys from the Parkes ''Murriyang'' radio telescope. A broad selection of simulated and injected expected signals (such as pulsars, fast radio bursts), poorly known signals (such as the features expected from flare stars) and unknown unknowns are generated for each survey. We provide a baseline by presenting how successful a typical pipeline based on the standard pulsar search software, PRESTO, is at finding the injected signals. The dataset is designed to aid in the development of new search algorithms, including image processing, machine learning, and deep learning. The raw data, ground truth labels, and baseline are provided. The collection is split into 4 parts. See collections in related links. Part 1 - Ground truth labels, injected images, multibeam dataset Part 2 - PAF dataset Part 3 - PAF dataset Part 4 - PAF dataset Publication: SPARKESX: Single-dish PARKES data sets for finding the uneXpected - A data challenge (Yong et a. 2022, submitted)

本研究介绍了面向未知异常探测的单口径帕克斯数据集（Single-dish PARKES Data Sets，SPARKESX），该数据集汇集了真实与模拟的高时间分辨率观测数据。
SPARKESX包含来自帕克斯“穆里扬”（Murriyang）射电望远镜的三次模拟巡天项目。
针对每一次巡天项目，我们均生成了多类经模拟与注入的预期信号（如脉冲星（pulsars）、快速射电暴（fast radio bursts））、认知尚不充分的信号（如耀星（flare stars）的预期辐射特征）以及未知未知信号。
本数据集提供了基准性能参考：基于标准脉冲星搜索软件PRESTO搭建典型处理流水线，测试其对注入信号的识别效果，以此作为基准性能基线。
该数据集旨在助力新型搜索算法的研发，涵盖图像处理、机器学习与深度学习等方向。
数据集提供原始数据、真实标签（ground truth labels）以及基准性能基线结果。
该数据集集合共分为4个子集，相关链接中可查阅各子集详情。
子集1：真实标签（ground truth labels）、注入图像与多波束数据集；子集2：PAF数据集；子集3：PAF数据集；子集4：PAF数据集
相关文献：SPARKESX：面向未知异常探测的单口径帕克斯数据集——一项数据挑战（Yong等，2022，已投稿）