Dataset · Hugging Face

maelic/RA-4M

Name: RA-4M
Creator: maelic
License: cc-by-nc-4.0
Keywords: task_categories:object-detection, language:en, license:cc-by-nc-4.0, size_categories:100K<n<1M, format:parquet, modality:image, modality:text, library:datasets, library:dask, library:polars, library:mlcroissant, arxiv:2609.12552, region:us, scene-graph, visual-relationship-detection

RA-4M — a verified free-text relation corpus 472,344 images · 4,282,531 relations · 10,102 free-text predicates · 497 object categories · 9.03 relations per image in the training split, plus a 24,964-image validation split (226,203 relations). Machine-annotated by an open-weight vision–language model, then filtered by deterministic geometric gates that reject 11.3% of raw candidates. Every object also carries a box-prompted SAM 3 mask. RA-4M exists because open-vocabulary… See the full description on the dataset page: https://huggingface.co/datasets/maelic/RA-4M.

View on Hugging FaceReport an issue

Author / organization	maelic
License (as declared)	cc-by-nc-4.0
Downloads	2.2K (as reported by Hugging Face)
Likes	5
Created	9 September 2026
Last updated at source	12 October 2026

scene-graphvisual-relationship-detectionopen-vocabularyrelation-detectioncoco-format

Embed this entry

<iframe src="https://silytics.si/embed/item/ddff99b0-8719-468f-a19f-7ed9bff2388e" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Popularity, last 2 days

2,183 now (+21 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.

More from maelic

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.