Jun 13Vibe with Hermes Agent — Bengaluru · RSVP
ToolsMCPBlogResearchCommunityStar on GitHub
All Tools
S
DataFreeOpen Source

SURYA

Multilingual OCR and document understanding with vision transformers

Apache-2.0

ABOUT

Extracting text from PDFs, scanned documents, and images reliably remains a hard problem for AI pipelines — especially with complex layouts, tables, math equations, and mixed-language content. Traditional OCR engines like Tesseract struggle with layout understanding and modern document structures. Surya solves this with vision-transformer models purpose-built for document understanding: it detects text regions, recognizes characters, analyzes page layout, extracts tables, and preserves reading order — all in a single pipeline with multilingual support for over 90 languages.

INSTALL
pip install surya-ocr

INTEGRATION GUIDE

1. Extract clean, structured text from scanned PDFs and images for RAG ingestion pipelines 2. Recognize and extract text from complex document layouts including tables, columns, and multi-language content 3. Process large batches of historical or multilingual documents with reading-order-preserving OCR 4. Extract table structures from scanned reports and financial documents for data analysis 5. Power automated document processing workflows that convert paper records into searchable digital archives

TAGS

ocrdocument-intelligencepdftext-detectionlayout-analysistable-recognitionmultilingualvision-transformer
Surya — AI Tool | Agentic AI For Good