Datasetpublic

EditFigure/real_layout_dataset_20260930

Real Layout Dataset (2026-09-30) This snapshot contains 47,767 source images and 24,301 JSON annotations. The images include 2,618 KBG images and 3,306 NeurIPS paper-figure images. Every JSON annotation has a source image with the same filename stem; not every image has an annotation. Files images_20260930.zip.part-000 through images_20260930.zip.part-022: consecutive pieces of one ZIP archive containing the images/ directory. images_20260930.zip.parts.sha256:… See the full description on the dataset page: https://huggingface.co/datasets/EditFigure/real_layout_dataset_20260930.

sourceHugging Faceupdated 9d agoView on Hugging Face

0likes403downloads

Dataset Card

Real Layout Dataset (2026-09-30)

This snapshot contains 47,767 source images and 24,301 JSON annotations. The images include 2,618 KBG images and 3,306 NeurIPS paper-figure images. Every JSON annotation has a source image with the same filename stem; not every image has an annotation.

Files

—images_20260930.zip.part-000 through images_20260930.zip.part-022: consecutive pieces of one ZIP archive containing the images/ directory.
—images_20260930.zip.parts.sha256: SHA-256 checksums for every piece and the reconstructed ZIP.
—json_20260930.zip: ZIP archive containing the json/ directory.

Both ZIP archives use AES-256 encryption. The extraction password is bitthu. The image archive is split for reliable transfer; the pieces are not separate datasets.

After downloading all pieces and the checksum file into one directory:

bash

cat images_20260930.zip.part-* > images_20260930.zip
sha256sum -c images_20260930.zip.parts.sha256
7z x -pbitthu images_20260930.zip
7z x -pbitthu json_20260930.zip

Image and annotation files correspond by filename stem. This is a snapshot of the cleaned annotation directory on 2026-09-30, not a claim that every image has been annotated or that it matches earlier Shaft training snapshots.