GutenOCR

### TL;DR

GutenOCR is an open-source vision-language model developed by Roots Automation, designed to enhance optical character recognition (OCR) by accurately grounding text to specific locations within documents. It addresses challenges in document workflows by providing precise text localization and understanding.

Key Insights & Metrics

Key Features

• Open-source code, weights, and data
• Unified, prompt-based interface for reading, detection, and grounding
• Trained on diverse datasets including business documents and scientific articles
• Supports full-page and localized reading with line- and paragraph-level bounding boxes
• Capable of handling conditional queries like 'where is x?'

Discussion