mmu_gz10 HATS Catalog Collection

This is the collection of HATS catalogs representing mmu_gz10.

This dataset is part of the Multimodal Universe, a large-scale collection of multimodal astronomical data. For full details, see the paper: The Multimodal Universe: Enabling Large-Scale Machine Learning with 100TBs of Astronomical Scientific Data.

Access the catalog

We recommend the use of the LSDB Python framework to access HATS catalogs. LSDB can be installed via pip install lsdb or conda install conda-forge::lsdb, see more details in the docs. The following code provides a minimal example of opening this catalog:

import lsdb

# Full sky coverage.
catalog = lsdb.open_catalog("https://huggingface.co/datasets/UniverseTBD/mmu_gz10")
# One-degree cone.
catalog = lsdb.open_catalog(
    "https://huggingface.co/datasets/UniverseTBD/mmu_gz10",
    search_filter=lsdb.ConeSearch(ra=134.0, dec=39.0, radius_arcsec=3600.0),
)

Each catalog in this collection is represented as a separate Apache Parquet dataset and can be accessed with a variety of tools, including pandas, pyarrow, dask, Spark, DuckDB.

File structure

This catalog is represented by the following files and directories:

• collection.properties � textual metadata file describing the HATS collection of catalogs
• mmu_gz10 � main HATS catalog directory
  • dataset/ � Apache Parquet dataset directory for the main catalog
    • ... parquet metadata and data files in sub directories ...
  • hats.properties � textual metadata file describing the main HATS catalog
  • partition_info.csv � CSV file with a list of catalog HEALPix tiles (catalog partitions)
  • skymap.fits � HEALPix skymap FITS file with row-counts per HEALPix tile of fixed order 10
• mmu_gz10_10arcs/ � default margin catalog to ensure data completeness in cross-matching, the margin threshold is 10.0 arcseconds
  • ... margin catalog files and directories ...

Catalog metadata

Metadata of the main HATS catalog, excluding margins and indexes:

Number of rows	Number of columns	Number of partitions	Size on disk	HATS Builder
17,736	7	766	2.6 GiB	hats-import v0.7.3, hats v0.7.3

Catalog columns

The main HATS catalog contains the following columns:

Name	_healpix_29	rgb_image	gz10_label	redshift	rgb_pixel_scale	ra	dec	object_id
Data Type	int64	struct<bytes: binary, path: string>	int32	float	float	double	double	string
Null count	0	N/A	0	0	0	0	0	0
Example row	359965024263830453	{'bytes': b'\x89PNG\r\n\x1a\n\x00\x00\x00\rIHDR\x00\x00\x01\x00\x00\x�	0	0.08629	0.262	133.7	39.42	521
Minimum value	11240848558758	N/A	0	-0.0001244	0.262	0.007269	-19.05	0
Maximum value	3458656132606848812	N/A	9	1.442	0.524	360	69.77	9999

Crossmatch with another catalog

HATS catalogs can be efficiently crossmatched using LSDB, which leverages the HEALPix partitioning to avoid loading the full datasets into memory:

import lsdb

mmu_gz10 = lsdb.open_catalog("https://huggingface.co/datasets/UniverseTBD/mmu_gz10")
other = lsdb.open_catalog("https://huggingface.co/datasets/<org>/<other_catalog>")

crossmatched = mmu_gz10.crossmatch(other, radius_arcsec=1.0)
print(crossmatched)

See the LSDB documentation for more details on crossmatching and other operations.

Dataset-specific context

Original survey
This dataset is based on Galaxy Zoo, a citizen science project where volunteers classify galaxy images according to their structure. The images are derived from the DESI Legacy Imaging Survey and correspond to what volunteers used for classification.

Data modality
The dataset includes RGB galaxy images (3×256×256) along with classification labels into 10 morphological classes. It also provides auxiliary tabular data such as right ascension (ra), declination (dec), redshift, and object identifiers.

Typical use cases
The dataset is mainly used for benchmarking and developing models for galaxy morphology classification, using clean and simplified labels. Several publications have used this dataset for evaluating different approaches (see examples).

Caveats
The dataset includes only a subset of Galaxy Zoo data with confident and clearly distinguishable labels. The images are RGB composites designed for visualization and classification, rather than full scientific measurements.

Citation
Users should acknowledge the Galaxy Zoo project and the DESI Legacy Imaging Surveys.