Direct Detection of White Dwarfs from DESI Legacy Imaging Surveys Images

Abstract

White dwarfs (WDs) represent the final evolutionary stage of low- to intermediate-mass stars. They are among the most abundant compact objects in the Milky Way. Large-scale WD surveys are therefore essential for probing stellar evolution and galactic structure. In recent years, large-scale digital sky surveys, such as the Sloan Digital Sky Survey and the Dark Energy Spectroscopic Instrument (DESI) Legacy Imaging Surveys, have amassed vast archives of photometric images. These datasets provide a rich resource for WD searches. We employ the WD Network to detect and localize WDs directly within 640 × 640 pixels photometric images. The framework enables fully automated identification of WD candidates from photometric images. On the test set, this model achieves a precision of 96.6%, a recall of 94.5%, and an F1 score of 95.5%. Images in the actual search are more diverse, and their distribution may differ from the test set, and these metrics may not directly reflect the model’s performance in the actual search. Using 125,872 photometric images from DESI Legacy Surveys Data Release 9, we detected 18,242 WD candidates. Cross-matching these candidates with DESI Data Release 1 spectra yielded 719 candidate spectra. We then performed manual visual inspections and confirmed 75 new WDs and 7 new CVs, which highlights substantial contamination among the remaining candidates with available spectra. We publicly released this WD candidates catalog online. These results expand the current census of known WDs and provide both a valuable data resource and a practical technical reference for future large-scale astronomical image processing.