Rendered English Wikipedia Articles as Images

A collection of the full English Wikipedia from February 1, 2018, rendered into images of 16x8464 resolution. Each rendered example contains a subset of one full Wikipedia article.

Use Cases

• Train a vision-language model on rendered Wikipedia article images to understand document layout and structure.
• Fine-tune an optical character recognition (OCR) system on the 16x8464 resolution images of English text.
• Analyze the visual representation of text from a specific Wikipedia dump date (February 1, 2018) for document analysis research.

Strengths

• Based on the full English Wikipedia from a specific date (February 1, 2018).
• Images are rendered at a consistent, high-resolution format of 16x8464 pixels.
• Articles underwent cleaning to strip markdown and unwanted sections like references.

Limitations

• Dataset is temporally stale, representing a snapshot from 2018 and not updated.
• The specific number of rows (images) and total dataset size are unknown.
• Each image contains only a subset of a full article, not the complete content.

Provenance

Source

English Wikipedia dump.

Collection Method

Built from a Wikipedia dump, with articles cleaned and rendered into images.

Time Range

Wikipedia snapshot from February 1, 2018.

Freshness

Last updated on the platform in August 2022, based on a 2018 Wikipedia dump.