Smith42/galaxies-with-hats

Galaxies with HATS

This is a HATS (HEALPix Adaptive Tiling Scheme) version of Smith42/galaxies (revision v2.0): ~8.5 million 256×256 pixel PNG galaxy cutouts from the DESI Legacy Survey DR8, centred on the galaxy source, spatially partitioned on the sky, and bundled with all ~160 metadata columns (Galaxy Zoo DESI morphologies, NSA photometry, redshifts, and OSSY / ALFALFA / JHU-MPA cross-matches) in every row.

Where the original dataset is ordered for ML training, this edition is ordered for sky-based analysis: every row carries ra/dec (degrees) and a _healpix_29 spatial index, and rows are partitioned into HEALPix tiles so that cone / box / cross-match queries touch only the relevant files.

The original 98/1/1 train/validation/test splits are preserved as three separate HATS collections:

Collection	Rows	HEALPix partitions	Size
train	8,474,566	1,555	~709 GB
validation	86,499	27	~7.3 GB
test	86,471	27	~7.3 GB

Stream it with 🤗 Datasets

HATS is plain Parquet, so the usual Hugging Face stack works out of the box (see Multimodal Universe × HATS):

from datasets import load_dataset

galaxies = load_dataset("Smith42/galaxies_with_hats", split="train", streaming=True)
example = next(iter(galaxies))
example["image"]  # -> PIL image, decoded automatically

Unlike the original ML-ordered repo, streaming order here follows the HEALPix sky tiling, meaning that consecutive examples are sky neighbours. Great for spatially coherent browsing; if you want i.i.d.-ish batches for training, add .shuffle(buffer_size=10_000) or stream the original Smith42/galaxies instead.

Sky queries with LSDB

We recommend LSDB (pip install lsdb) for spatial work with HATS catalogs:

import lsdb

# Open a split (lazy -- nothing is downloaded yet)
galaxies = lsdb.open_catalog("hf://datasets/Smith42/galaxies_hats/train")

# One-degree cone around (RA, Dec) = (161.0, 48.0)
cone = lsdb.open_catalog(
    "hf://datasets/Smith42/galaxies_hats/train",
    search_filter=lsdb.ConeSearch(ra=161.0, dec=48.0, radius_arcsec=3600.0),
)
df = cone.compute()  # materialise to pandas; df["image"]["bytes"] holds the JPGs

# Cross-match your own catalog against the galaxy images
# my_cat = lsdb.from_dataframe(my_dataframe, ra_column="ra", dec_column="dec")
# matched = my_cat.crossmatch(galaxies, radius_arcsec=1.0)

Each catalog is also a plain Apache Parquet dataset (<split>/<split>/dataset/), directly readable with pandas, pyarrow, dask, Spark, or DuckDB — partitions are laid out as Norder={order}/Dir={dir}/Npix={pixel}.parquet HEALPix tiles of ~10,000 rows (≈0.8 GB) each.

import pandas as pd  # read a single sky tile

tile = pd.read_parquet(
    "hf://datasets/Smith42/galaxies_hats/test/test/dataset/Norder=0/Dir=0/Npix=4.parquet"
)

Catalog columns

All three collections share an identical 170-column schema: the original image (struct of PNG bytes + path, decoded as an image by 🤗 datasets), dr8_id, galaxy_size, ra, dec, and all metadata columns of Smith42/galaxies v2.0, plus the _healpix_29 spatial index added by HATS.

The main HATS catalog contains the following columns:

Name	_healpix_29	image	dr8_id	galaxy_size	ra	dec	brickid	objid	smooth-or-featured_smooth_fraction	smooth-or-featured_featured-or-disk_fraction	smooth-or-featured_artifact_fraction	disk-edge-on_yes_fraction	disk-edge-on_no_fraction	has-spiral-arms_yes_fraction	has-spiral-arms_no_fraction	bar_strong_fraction	bar_weak_fraction	bar_no_fraction	bulge-size_dominant_fraction	bulge-size_large_fraction	bulge-size_moderate_fraction	bulge-size_small_fraction	bulge-size_none_fraction	how-rounded_round_fraction	how-rounded_in-between_fraction	how-rounded_cigar-shaped_fraction	edge-on-bulge_boxy_fraction	edge-on-bulge_none_fraction	edge-on-bulge_rounded_fraction	spiral-winding_tight_fraction	spiral-winding_medium_fraction	spiral-winding_loose_fraction	spiral-arm-count_1_fraction	spiral-arm-count_2_fraction	spiral-arm-count_3_fraction	spiral-arm-count_4_fraction	spiral-arm-count_more-than-4_fraction	spiral-arm-count_cant-tell_fraction	merging_none_fraction	merging_minor-disturbance_fraction	merging_major-disturbance_fraction	merging_merger_fraction	file_name	iauname	est_petro_th50	mag_r_desi	mag_g_desi	mag_z_desi	petro_theta	petro_th50	petro_th90	petro_phi50	petro_phi90	petro_ba50	petro_ba90	elpetro_ba	elpetro_phi	elpetro_flux_r	elpetro_theta_r	elpetro_mass	sersic_n	sersic_ba	sersic_phi	elpetro_absmag_f	sersic_nmgy_f	elpetro_absmag_n	sersic_nmgy_n	elpetro_absmag_u	sersic_nmgy_u	elpetro_absmag_g	sersic_nmgy_g	elpetro_absmag_r	sersic_nmgy_r	elpetro_absmag_i	sersic_nmgy_i	elpetro_absmag_z	sersic_nmgy_z	redshift_nsa	mag_f	mag_n	mag_u	mag_g	mag_r	mag_i	mag_z	u_minus_r	elpetro_mass_log	dr7objid_ossy	ra_ossy	dec_ossy	redshift_ossy	log_l_oiii	fwhm	e_fwhm	equiv_width	log_l_ha	log_m_bh	upper_e_log_m_bh	lower_e_log_m_bh	log_bolometric_l	ra_alf	dec_alf	W50	sigW	W20	HIflux	sigflux	SNR	RMS	Dist	sigDist	logMH	siglogMH	ra_jhu	dec_jhu	fibre_sfr_avg	fibre_sfr_entropy	fibre_sfr_median	fibre_sfr_mode	fibre_sfr_p16	fibre_sfr_p2p5	fibre_sfr_p84	fibre_sfr_p97p5	fibre_ssfr_avg	fibre_ssfr_entropy	fibre_ssfr_median	fibre_ssfr_mode	fibre_ssfr_p16	fibre_ssfr_p2p5	fibre_ssfr_p84	fibre_ssfr_p97p5	total_ssfr_avg	total_ssfr_entropy	total_ssfr_flag	total_ssfr_median	total_ssfr_mode	total_ssfr_p16	total_ssfr_p2p5	total_ssfr_p84	total_ssfr_p97p5	total_sfr_avg	total_sfr_entropy	total_sfr_flag	total_sfr_median	total_sfr_mode	total_sfr_p16	total_sfr_p2p5	total_sfr_p84	total_sfr_p97p5	photoz_id	ra_photoz	dec_photoz	photo_z	photo_zerr	spec_z	mag_abs_g_photoz	mag_abs_r_photoz	mag_abs_z_photoz	mass_inf_photoz	mass_med_photoz	mass_sup_photoz	sfr_inf_photoz	sfr_sup_photoz	ssfr_inf_photoz	ssfr_med_photoz	ssfr_sup_photoz	sky_separation_arcsec_from_photoz	redshift	est_petro_th50_kpc	__index_level_0__
Data Type	int64	struct<bytes: binary, path: string>	string	int64	double	double	int32	int32	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	string	string	double	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	float	double	double	float	float	float	float	float	float	float	double	float	float	float	float	float	float	float	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	double	int64
Null count	0	N/A	0	0	0	0	0	0	0	0	0	7,585,145 (89.5%)	7,585,145 (89.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	7,923,531 (93.5%)	1,508,220 (17.8%)	1,508,220 (17.8%)	1,508,220 (17.8%)	8,288,807 (97.81%)	8,288,807 (97.81%)	8,288,807 (97.81%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	8,116,934 (95.78%)	0	0	0	0	0	7,882,536 (93.01%)	0	0	14,543 (0.17%)	17,007 (0.2%)	7,882,536 (93.01%)	7,882,536 (93.01%)	7,882,536 (93.01%)	7,882,571 (93.01%)	7,882,570 (93.01%)	7,882,571 (93.01%)	7,882,570 (93.01%)	7,882,570 (93.01%)	7,882,570 (93.01%)	7,882,536 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,536 (93.01%)	7,882,536 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,537 (93.01%)	7,882,542 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,536 (93.01%)	7,882,542 (93.01%)	7,882,540 (93.01%)	7,882,536 (93.01%)	7,934,716 (93.63%)	7,902,017 (93.24%)	7,882,672 (93.02%)	7,882,572 (93.01%)	7,882,571 (93.01%)	7,882,596 (93.01%)	7,882,795 (93.02%)	7,882,579 (93.01%)	7,882,542 (93.01%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,471,529 (99.96%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	8,446,786 (99.67%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,440 (90.88%)	7,701,440 (90.88%)	7,701,438 (90.88%)	7,701,440 (90.88%)	7,701,440 (90.88%)	7,701,440 (90.88%)	7,701,440 (90.88%)	7,701,438 (90.88%)	7,701,686 (90.88%)	7,702,153 (90.89%)	7,701,438 (90.88%)	7,702,153 (90.89%)	7,702,153 (90.89%)	7,702,153 (90.89%)	7,702,153 (90.89%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	7,701,438 (90.88%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	632,529 (7.46%)	624,776 (7.37%)	7,400,771 (87.33%)	624,776 (7.37%)	626,277 (7.39%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	624,776 (7.37%)	620,954 (7.33%)	620,954 (7.33%)	0
Example row	396125208651480696	{'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x01\x00\x00\x…	594276_229	4010	179.5	52.64	594276	229	0.7793	0.1612	0.05948													0.09985	0.8888	0.01132													0.7796	0.1714	0.03812	0.01085	data/594276_229.jpg		2.622	17.73	18.19	17.4																																																																																																						8.002e+15	179.5	52.64	0.07962	0.0192		-19.72	-20.14	-20.45	8.916	8.95	8.984	1.732	1.868	-7.196	-7.128	-7.044	0	0.07962	3.943	7154494
Minimum value	739458926660	N/A	100000_1081	1094	2.953e-05	-65.9	28612	0	0.01351	0.01837	0.02748	0.01325	0.01119	0.02962	0.01014	0.01488	0.02545	0.02917	0.006299	0.008042	0.02487	0.017	0.006605	0.01036	0.01239	0.007499	0.01565	0.01061	0.02994	0.01303	0.02092	0.01085	0.009553	0.02966	0.006751	0.006136	0.006134	0.01004	0.03816	0.02309	0.01019	0.008846	data/100000_1081.jpg	J000000.74-091320.2	0.8155	8.82	-13.78	-16.47	-0	-0	-0	0.01331	0.007782	-0.9724	-0.9724	-0.9724	0.007782	-1.958e+05	-9999	-0	0.5	0.15	0.0003052	-26.29	-1.794e+15	-26.32	-1.832e+16	-27.26	-1.15e+04	-28.27	-2.413e+07	-28.59	-1.538e+07	-28.64	-3.35e+07	-28.14	-4.193e+07	-0.003335	-15.01	-3.218	7.685	8.314	5.779	6.685	7.777	-7.035	-∞	5.877e+17	0.5123	-10.96	0.01596	-9999	1320	4.1	8.5	40.59	6.13	-0	-0	-9.99	0.00333	-0.2061	9	1	-0	-7.61	0.02	1.2	1.28	2.1	0.2	6.11	0.04	0.000686	-11.25	-5	-6.917	-∞	-5	-∞	-∞	-∞	-∞	-13.3	-6.957	-13.35	-16	-13.79	-14.87	-12.87	-12.27	-99	-7.253	-0	-99	-99	-99	-99	-99	-99	-99	-6.851	-0	-99	-99	-99	-99	-99	-99	8e+15	2.953e-05	-65.89	0.001	3.075e-05	0.0001144	-32.36	-32.48	-33.13	-99	-99	-99	-99	-99	-99	-99	-99	-0	-0.003335	-0.9113	0
Maximum value	3458764069707388219	N/A	99999_704	249824	360	84.78	660728	31041	0.9204	0.9551	0.9179	0.9888	0.9868	0.9899	0.9704	0.8959	0.6597	0.9578	0.5212	0.6231	0.8273	0.9148	0.8917	0.9714	0.965	0.9764	0.6663	0.9517	0.9707	0.9373	0.731	0.9654	0.911	0.9506	0.8075	0.564	0.5158	0.7038	0.9512	0.5815	0.7444	0.8767	data/99999_704.jpg	J235959.50+003645.4	20.16	19	∞	∞	254.8	154.4	163.4	180	180	1	1	1	180	6.528e+05	357.4	5.242e+12	6	1	180	∞	1.012e+15	∞	1.937e+10	∞	8.431e+05	∞	4.725e+05	∞	4.879e+06	∞	2.119e+06	∞	7.75e+05	0.15	∞	∞	∞	∞	∞	∞	∞	12.16	12.72	5.888e+17	359.8	68.5	0.1999	43.22	1.556e+04	4.29e+04	833.9	43.36	9.09	1.68	1.09	46.77	360	36.28	885	296	896	424.3	1.16	833	9.74	259.6	30.9	10.94	0.81	360	84.27	3.95	0	3.925	3.95	3.908	3.901	3.942	3.949	-1	0	-1	-1	-1	-1	-1	-1	-6.028	0	3	-6.037	-6	-6.095	-6.205	-6.012	-6.002	3.9	0	2	3.875	3.9	3.858	3.851	3.892	3.899	8.002e+15	360	84.77	0.5	0.2996	2.2	21.26	-10.01	19.15	13.17	13.2	13.24	5.824	5.965	-7.068	-7	-6.932	9.982	2.2	102.8	8689368

Full per-collection details (file structure, partition lists, per-column tables) are in each collection's own README: train, validation, test.

Provenance and fidelity

• Source: Smith42/galaxies, revision v2.0 (images and metadata pre-merged), CC-BY-SA-4.0.
• Conversion: hats-import v0.9.2 / hats v0.9.2, partitioned on ra/dec with pixel_threshold=10_000, sorted by dr8_id within partitions.

The original dataset

Here we have ~8.5 million galaxy cutouts from the DESI legacy survey DR8, centred on the galaxy source (256×256 pixel PNGs in the v2.0 revision used here; the original main revision serves 512×512 JPGs). 1% of the images are split into a test set, and 1% into a validation set; the remaining 98% comprise the training set.

For the ML-ordered original and AstroPT embeddings, see Smith42/galaxies and Smith42/galaxies_embeddings.

Useful links

Paper here: https://arxiv.org/abs/2405.14930

Models here: https://huggingface.co/Smith42/astroPT

And here: https://huggingface.co/Smith42/astroPT_v2.0

Code here: https://github.com/smith42/astroPT

Upstream catalogue is on Zenodo and the paper describing the catalogue is available as Walmsley+2023.

HATS and LSDB are developed by the LINCC Frameworks team — see lsdb.io for citation guidance.

If you find this dataset useful please consider citing the sources below 🚀🚀:

@article{ref_dey2019,
    author = {Dey, A. and Schlegel, D. J. and Lang, D. and Blum, R. and Burleigh, K. and Fan, X. and Findlay, J. R. and Finkbeiner, D. and Herrera, D. and Juneau, S. and others},
    title = {{Overview of the DESI Legacy Imaging Surveys}},
    journal = {Astronomical Journal},
    volume = {157},
    number = {5},
    pages = {168},
    year = {2019},
    issn = {1538-3881},
    publisher = {The American Astronomical Society},
    doi = {10.3847/1538-3881/ab089d}
}

@article{ref_walmsley2023,
    author = {Walmsley, M. and G{\ifmmode\acute{e}\else\'{e}\fi}ron, T. and Kruk, S. and Scaife, A. M. M. and Lintott, C. and Masters, K. L. and Dawson, J. M. and Dickinson, H. and Fortson, L. and Garland, I. L. and others},
    title = {{Galaxy Zoo DESI: Detailed morphology measurements for 8.7M galaxies in the DESI Legacy Imaging Surveys}},
    journal = {Monthly Notices of the Royal Astronomical Society},
    volume = {526},
    number = {3},
    pages = {4768--4786},
    year = {2023},
    issn = {0035-8711},
    publisher = {Oxford Academic},
    doi = {10.1093/mnras/stad2919}
}

@article{ref_smith2024,
    author = {Smith, M. J. and Roberts, R. J. and Angeloudi, E. and Huertas-Company, M.},
    title = {{AstroPT: Scaling Large Observation Models for Astronomy}},
    journal = {ArXiv e-prints},
    year = {2024},
    eprint = {2405.14930},
    doi = {10.48550/arXiv.2405.14930}
}