metadata

annotations_creators: []
language: []
language_creators: []
license: []
multilinguality: []
pretty_name: Internet Archive historic book pages unlabelled.
size_categories: []
source_datasets: []
tags: []
task_categories: []
task_ids: []

Data card for Internet Archive historic book pages unlabelled.

10,844,387 unlabelled pages from historical books from the internet archive.
Intended to be used for:
- pre-training computer vision models in an unsupervised manner
- using weak supervision to generate labels