Skip to content

Dataset · JBrightmanAI

Millions of PDF Document Pages for VLM Models

JBrightmanAI/pdfa-eng-wds

A massive corpus of filtered English PDF documents with layout annotations for training multimodal models.

What's inside

The dataset contains 10 to 100 million English records collected from the SafeDocs corpus (CC-MAIN-2021-31-PDF-UNTRUNCATED). The data is specifically filtered and prepared for machine learning, including added bounding boxes on document pages.

What you can do with it

  • Pre-training and fine-tuning vision-language models (VLM) for layout and text understanding in documents
  • Training page element detection and complex PDF parsing systems
  • Training OCR pipelines on diverse real-world documents
Tasks
Image → text
Languages
English
Size
10M–100M rows
Downloads this month
55.4K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("JBrightmanAI/pdfa-eng-wds")