Published November 12, 2025 | Version 1.0

Augments of local HST galaxy images at different noise and resolution levels

Description

Description

The dataset contains 189 local galaxies from the RNGC/IC catalog that are imaged by the Hubble Space Telescope in the F814W filter. For each galaxy, we created a series of augments by degrading the spatial resolution of the image and adding noise to decrease the signal-to-noise, or the surface brightness limit. This produced a dataset of ~60k images at varying noise and resolution levels. We used these images to investigate the dependence of common structural measurements on resolution and noise. We plan to use these further to train a deep learning network that is unbiased to imaging effects. The dataset and the project are described in detail in Sazonova et al. (2025). 

Please reach out to liza.sazonova@uwaterloo.ca if you would like to learn more about the dataset, and please reference our paper if you wish to use it in your projects. 

Contents

Filename Description
data.tar.zst

A compressed archive containing all the augments. For each galaxy, we generate images at resolutions of 25 to 2000 parsecs/px, and surface brightness limits ranging from 20 to 26 mag/arcsec2. Each galaxy then has up to 450 augments.

This is an archive containing all augments for all galaxies, i.e. ~60,000 images. Th file contains 190 tar.zst files for each galaxy. Each of these contains ~450 fits.fz compressed files. The total size of a decompressed archive is ~350 GB.

segmaps.json.gz Segmentation maps and masks, outlining each source and any foreground/background contaminants, for each augmented image. The segmentation maps are saved in the RLE format. Please consult our package for help in converting these to boolean masks.
sample.csv The list of 190 galaxies in this study and their properties, such as coordinates, radii, distances, etc.
morphology.csv Morphological measurements obtained with statmorph-lsst and Galfit for each augmented image.

Files

morphology.csv

Files (19.6 GB)

Name Size
md5:0df9202bcb8a6f4eb8f3db2786ceb92d
19.4 GB Download
md5:a8e645ba46b4cdb45aeaba274947918b
90.8 MB Preview Download
md5:9ecfaa319258feee4b33a07cae262719
24.6 kB Preview Download
md5:51793b9d601d05943df48d809a332186
87.1 MB Download

Additional details

Dates

Created
2025-11-11

Software

Repository URL
https://github.com/astro-nova/statmorph-lsst
Programming language
Python