Command Palette
Search for a command to run...
Britannica Illustrated Pages Dataset
Britannica Illustrated Pages is published by BigLAM for researching the distribution of illustrations and image content across editions of the Encyclopaedia Britannica, offering a large-scale resource of annotated images.
The data is sourced from scanned volumes of the Encyclopaedia Britannica, from the 1st edition (1768–1771) to the 14th edition (1929). Through a page classifier, 115,293 pages containing illustrations were selected from 975,345 pages of raw material.
Dataset Composition
- pages subset: Contains 115,293 entries, storing scaled 600 px JPEG illustration pages along with complete metadata.
- manifest subset: Covers all 975,345 pages of the original volumes, recording classifier scores, OCR text statistics, and source provenance information, without images.
- crop_masks subset: Generated based on RF‑DETR Seg model predictions, including 115,293 pages with a total of 700,480 illustration instances (confidence ≥ 0.10), providing bounding boxes and segmentation masks for illustration instances.
Core fields:
- ia_id
- volume
- page (source volume location identifier)
- p_illustrated: probability of illustration detection
- words: number of OCR-recognized words
- stratum: page type
- tier
- bucket_url
- file_key
Citation
When using this dataset, please cite the Internet Archive source of the corresponding page (ia_url) and the contributing institution of the scanned copy (if applicable). Training labels were provided by the National Library of Scotland; the classifier, scanning, and data curation work was carried out by Daniel van Strien, and the dataset is published through BigLAM.
Build AI with AI
From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.