Dataset of artefacts and nominal objects for machine learning applications in astronomy

We present two datasets, each containing 1127 image cutouts from ZTF DR3:

Artefact dataset – images containing artefacts; isolated using the active anomaly detection algorithm PineForest, developed by the SNAD team.

Nominal dataset – consisting of normal objects selected from the same ZTF fields as the artefact dataset.

Data Format

Both datasets are provided in FITS format in two sizes:

28 × 28 pixels

63 × 63 pixels

FITS Headers

Each FITS file includes the following metadata in its header in addition to standard ZTF headers:

OID – ZTF Object ID

OIDRA – Object Right Ascension (deg)

OIDDEC – Object Declination (deg)

TAGS – Labels for the central object according to the SNAD classification schema (only for artefact dataset)

URL – Download URL for the FITS image

These datasets are intended for use in machine learning training, anomaly detection studies, and astrophysical catalog cleaning. If you use this dataset, please cite the corresponding SNAD publications.

For more information, visit: https://snad.space/