PaddleOCR: Project Introduction
PaddleOCR converts PDF documents and images into JSON/Markdown and is released under the Apache 2.0 license.
Verified
Core facts
Read by a script from the official API, not rewritten by a model.
- License
- Apache-2.0
- Languages
- Python
- Maintainer
- PaddlePaddle
- Homepage
- https://www.paddleocr.com
- Verified
What it is
PaddleOCR converts PDF documents and images into structured, LLM-ready data (JSON/Markdown). This project is released under the Apache 2.0 license. PaddleOCR-VL-0.9B is a vision-language model (VLM) that integrates a NaViT-style dynamic resolution visual encoder with the ERNIE-4.5-0.3B language model. The project says HPD-Parsing is a lightweight vision-language model designed for high-throughput document parsing.
Main features
PP-StructureV3 converts complex PDFs and images into Markdown or JSON.
Unlike the PaddleOCR-VL series models, PP-StructureV3 provides more fine-grained coordinate information, including table cell coordinates and text coordinates.
PaddleOCR-VL supports 109 languages.
The project reports that PP-OCRv6 supports 50 languages with a single unified model, including Chinese, English and Japanese.
PaddleOCR supports various hardware backends (NVIDIA GPU, Intel CPU, Kunlunxin XPU, and diverse AI accelerators).
PaddleOCR is deeply integrated with Dify, RAGFlow, Pathway, and Cherry Studio.
Getting started
The PaddleOCR official website provides an interactive Experience Center and APIs.
For local usage, the project asks readers to refer to the documentation that matches their needs.
Models can be converted to ONNX format.
Inference can be accelerated using multi-GPU and multi-process.
PaddleOCR can be integrated into applications written in C++, C#, Java, and similar languages.
Who it is for
20 major models support Transformers as the inference backend. PaddleOCR converts common document formats such as Word, Excel, and PowerPoint into Markdown. The PaddleOCR-VL series, PP-StructureV3, and PP-DocTranslation support exporting parsed results to DOCX.
PaddleOCR.js, the official browser inference SDK, supports running PP-OCRv5 directly in the browser. HPD-Parsing supports both OpenAI-compatible serving and local inference through a customized vLLM runtime. The PP-OCRv6 tiers tiny (1.5M), small (7.7M), and medium (34.5M) are for edge, mobile, and server deployment.
The project reports that the PP-OCRv6 medium tier achieves +4.6% detection and +5.1% recognition over PP-OCRv5_server. PP-OCR series models support returning single-character coordinates. PaddleOCR supports chart-to-table conversion via the PP-Chart2Table module.
PaddleOCR adds AIStudio, ModelScope, and other model download sources, and users can specify the source for model downloads. PaddleOCR supports NVIDIA RTX 50 series graphics cards on Windows. The PP-OCRv5 C++ local deployment solution has been comprehensively upgraded and now supports both Linux and Windows.
High-performance inference supports CUDA 12, and inference can be performed using either the Paddle Inference or ONNX Runtime backends.
Related tools
Related tools listed on this site.
百灵大模型
Ant Ling is a language and multimodal model family. Official documentation describes online multi-turn conversations, text generation/content creation, API access and cross-modal understanding/generation.
FlagEval
Evaluation toolkit and open platform for large models. Its public project covers foundation models, training and fine-tuning algorithms, with language, vision, audio and multimodal evaluation directions.
OpenCompass
Model evaluation platform with language and multimodal leaderboards, benchmark resources, and documentation.
MMBench
An OpenCompass Community multimodal-understanding benchmark using multiple-choice questions and fine-grained ability dimensions to analyze vision-language models, with Chinese and English data and evaluation-tool links.
图改改
TuGaiGai is an online editor focused on text within images. Its official page describes changing, adding and removing image text, font recognition, image translation and image/PDF conversion.
多维视界
Duowei Shijie is an audiovisual analysis tool whose official site describes transcription, translation, summaries, interactive learning and content rewriting.
Winston AI
AI-content detection and plagiarism-checking tool whose site describes text scans, OCR input, sentence prediction maps and reports.
文心大模型
ERNIE (Wenxin large model) is Baidu's large-model family; its official ERNIE Blog publishes ERNIE 5.x and other models, the ERNIE-Image text-to-image model and related papers, and links to ERNIE Bot, the Qianfan platform and AI Studio.
Sources
This page is written from the official sources below. Open them to check the original text.
- GitHub READMEhttps://github.com/PaddlePaddle/PaddleOCRFetched
Verified ·
This page is written by AI from the official sources above and checked automatically. This site has not tested it. Found a mistake? Tell us.