Supernova Detection Datasets

Supernovae are dying stars whose light curves change in certain special patterns, which helps astronomers infer the expansion coefficient of the Universe. A series of sky surveys were launched in search of supernovae and generated tremendous amount of data, which pushed astronomy into a new era of big data. While the traditional machine learning methods performs well to deal with such data, the deep learning methods such as convolutional neural networks demonstrate more powerful adaptability for big data in this area. However, most data in the existing works are either simulated or without generality.
To address these problems, we collected and sorted all the known objectives of the Pan-STARRS and the Popular Supernova Project (PSP) and produced two datasets and then compared the YOLOv3 and the FCOS algorithm on them.