yunfeichen/world-bird-species

World Bird Species (IOC 15.2)

One record for each of the 11,064 living bird species in the IOC World Bird List v15.2 (44 orders, 250 families; the 163 extinct species are left out).

This is the data half of huashu-bird-sticker, a project that also drew one sticker per species. The stickers themselves live in the GitHub repo; this mirror carries only the tables, so you can load_dataset them without pulling ~700 MB of images.

Files

File	Contents
species.csv (default config)	Taxonomy (IOC sequence, order, family), scientific name and authority, English and Chinese names, Wikidata / eBird / GBIF / iNaturalist IDs, sticker file path, sticker review score and whether it meets the quality bar, and the reference photo each sticker was drawn from
names.csv (names config)	Common names in 43 languages (46 columns), from the IOC multilingual list
species.json	The complete records, including ranges, subspecies, AVONET traits (mass, wing, tail, bill, habitat, migration, trophic niche), IUCN and CITES status
languages.json	Language codes and the IOC column each one comes from
FIELDS.md	Field-by-field description of species.json, with sources and coverage
NOTICE.md	Third-party sources and their licenses

from datasets import load_dataset
species = load_dataset("yunfeichen/world-bird-species", "species", split="train")
names = load_dataset("yunfeichen/world-bird-species", "names", split="train")

Coverage

• Names: English covers every species; French and Simplified Chinese nearly all; Japanese, Traditional Chinese, Russian, German, Spanish, Polish, Dutch, Swedish and about a dozen more European languages over 90%. Korean, Arabic, Thai, Persian and some others cover a few hundred to about a thousand species, because that is what the IOC list has.
• IDs: Wikidata 10,939, GBIF 10,649, eBird 10,052.
• AVONET traits: 10,577 species. IUCN category: 10,180.

About the sticker columns

11,030 stickers are AI illustrations drawn from a reference photo (or a board of several images) and scored by a separate AI reviewer against that reference; 34 are cut out from real photographs. meets_quality_bar is true for 7,676 species and false for 3,388. Scores come from a model, not an ornithologist. See the GitHub README for how they were made.

Licenses

• Data compiled by this project: MIT.
• Taxonomy and multilingual names: IOC World Bird List v15.2, Gill, Donsker & Rasmussen (Eds), CC BY 3.0.
• Traits: AVONET (Tobias et al. 2022, Ecology Letters, doi:10.1111/ele.13898), CC BY 4.0.
• Wikidata-derived fields: CC0.
• The stickers referenced by sticker_file are not in this dataset; they are under PolyForm Shield 1.0.0 in the GitHub repo.

Full details in NOTICE.md. Corrections welcome as issues on the GitHub repo.