Astronet-Triage-v2 dataset

Dataset created for the production of Astronet-Triage-v2, a CNN that classifies eclipsing signals (eg from exoplanetary transits or eclipsing binary star systems) from other types of signals. The dataset contains ~25,000 signals, detected via Box Least Squares (Kovacs 2002) in the first 3 years of the TESS mission. Each signal was 1 of 5 labels by at least 3 individuals. Most conflicts were resolved into a consensus label for that signal. The dataset contains 1) a csv of all signals (threshold-crossing events, or TCEs) and 2) a tar of QLP light curve data used. Both are used as inputs to Astronet-Triage-v2, which can be found here: https://github.com/mdanatg/Astronet-Triage/tree/265366bc9e8e98d248df34fad267704324c344ce本数据集专为开发Astronet-Triage-v2而创建，该模型是一种卷积神经网络（Convolutional Neural Network，CNN），用于从各类信号中区分出凌日信号（例如系外行星凌日或食双星系统产生的信号）。本数据集包含约25000个信号，这些信号是在TESS任务前三年通过箱型最小二乘法（Box Least Squares，Kovacs 2002）检测得到的。每个信号均由至少3名标注人员标注为5个类别之一。绝大多数标注冲突已被协调为该信号的共识标注。本数据集包含两部分内容：1）所有信号（阈值穿越事件，threshold-crossing events，TCEs）的csv文件；2）所用QLP光变曲线数据（Quick Look Pipeline light curve data）的tar归档文件。二者均作为Astronet-Triage-v2的输入数据，该模型的开源地址为：https://github.com/mdanatg/Astronet-Triage/tree/265366bc9e8e98d248df34fad267704324c344ce