Repolode
Submit
PaddlePaddle

PaddlePaddle/PaddleOCR

Open-source OCR and document AI toolkit that converts PDFs and images into structured Markdown or JSON for LLM and RAG pipelines.

Open-source alternative to:Adobe AcrobatABBYY FineReader

View on GitHubWebsite
Stars
92,370
Forks
11,626
Open issues
252
Last commit
2 months ago

What is it?

PaddleOCR is an OCR toolkit from PaddlePaddle that turns PDF documents and images into structured output such as Markdown and JSON, aimed at feeding content into large language models and retrieval-augmented generation pipelines. It covers both document parsing, which preserves layout, tables, formulas and charts, and scene text recognition for multilingual text in photos and street views.

The project offers several model families. PP-OCR models handle text detection and recognition across 100+ languages, while PaddleOCR-VL is a 0.9B vision-language model for document parsing. PP-StructureV3 provides finer-grained coordinate information, including table cell and text positions. Models can run on CPU, NVIDIA GPU, Intel, Kunlunxin XPU and other accelerators, and can be exported to ONNX or run through OpenVINO, ONNX Runtime or TensorRT.

For deployment, PaddleOCR provides a hosted experience center and API, local Python installation, and a serving option for integration from C++, C#, Java and other languages.

Key features

  • 01Converts PDFs and images into Markdown or JSON output.
  • 02PaddleOCR-VL 0.9B vision-language model for document parsing.
  • 03Text detection and recognition across 100+ languages with a unified model.
  • 04Layout analysis that retrieves table cell and text coordinates.
  • 05Runs on CPU, NVIDIA GPU, Intel, Kunlunxin XPU and other accelerators.
  • 06Export to ONNX and acceleration via OpenVINO, ONNX Runtime or TensorRT.

Use cases

  • Preparing scanned PDFs and reports as clean Markdown for a RAG knowledge base.
  • Extracting tables and text with coordinates for invoice or form processing.
  • Recognizing multilingual text in photos, signs or industrial components.
  • Integrating OCR into a C++, C# or Java application through the serving option.

In collections

PDF, OCR & document toolsComputer vision libraries & tools

Topics

#ocr#document-parsing#pdf-parser#rag#computer-vision#vision-language-model#multilingual#deep-learning#ai4science#chineseocr

FAQ

What is PaddleOCR?
Open-source OCR and document AI toolkit that converts PDFs and images into structured Markdown or JSON for LLM and RAG pipelines. PaddleOCR is an OCR toolkit from PaddlePaddle that turns PDF documents and images into structured output such as Markdown and JSON, aimed at feeding content into large language models and retrieval-augmented generation pipelines. It covers both document parsing, which preserves layout, tables, formulas and charts, and scene text recognition for multilingual text in photos and street views.
Is PaddleOCR free to use? What license does it have?
PaddleOCR is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is PaddleOCR still maintained?
Updated within 6 months. The last commit to PaddleOCR was on Sep 16, 2026.
What can I use PaddleOCR for?
Preparing scanned PDFs and reports as clean Markdown for a RAG knowledge base. Extracting tables and text with coordinates for invoice or form processing. Recognizing multilingual text in photos, signs or industrial components. Integrating OCR into a C++, C# or Java application through the serving option.
Where can I get PaddleOCR?
The source code and installation instructions are on GitHub: https://github.com/PaddlePaddle/PaddleOCR

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects