PaddleOCR

PaddleOCR is an OCR and document parsing toolkit for developers turning PDFs and images into data for document search, RAG systems and AI agents. Its local models cover both structured pages and text found in photographs.

• Document parsing: PaddleOCR-VL reads text, tables, formulas and charts, including difficult material such as ancient documents, rare characters and seals, and produces Markdown or JSON.
• Layout and coordinates: PP-Structure converts complex documents into structured output with detailed positions for text and individual table cells.
• Multilingual recognition: PP-OCR recognizes Chinese, English, Japanese and Latin-script languages within one model, and supports text in street scenes, books, IDs and industrial components.
• Application integration: Developers can use the toolkit in C++, C# and Java applications, export models to ONNX, and accelerate inference with OpenVINO, ONNX Runtime or TensorRT.

The Python project is open source under Apache 2.0 and supports local deployment, CPU inference and parallel processing across multiple GPUs. The official website also offers a separate hosted interface and API for online processing.