Deep learning on galaxy images: training dataset

This page is used to collect the training dataset for deep learning applications on galaxy images.

If you want to contribute, please contact Shiyin Shen (ssy@shao.ac.cn)

local galaxy sample

• galaxy zoo data [1]

• Galaxy Zoo DECaLS[2]
• stamps of SDSS main sample galaxies(r<17.77) in DESI [3]

high z galaxy sample

• COSMOS F814W galaxy postages
• CANDELS 5 fields galaxy postages (H<24.5)

Mock galaxies

• generate using GalSim (with CSST-like survey parameters)

• tel_diam=200mm, exptime=150s, lam=5500A, sky_mag=23, psf_sigma=0.2, gain=1, readnoise=5, pixel_scale=0.074

• generate using GalSim-Hub[4]

• A sample of CGS galaxies{https://arxiv.org/abs/1111.4605] manually move to high redshifts and observed in other surveys(PS1, SDSS stripe82, CANDELS_AEGIS , Hubble Frontier Fields

• The Carnegie-Irvine Galaxy Survey (CGS) is a long-term program to investigate the photometric and spectroscopic properties of a statistically complete sample of 605 bright (BT < 12.9mag), southern (δ < 0°) galaxies using the facilities at Las Campanas Observatory.
• contact Shiyin Shen for more infomation