PaddleOCR
2.9.1
English | Simplified Chinese
PaddleOCR aims to create a rich, leading and practical OCR tool library to help developers train better models and implement them.
PaddleOCR is overseen by PMC. Issues and PRs will be reviewed on a best-effort basis. For a complete overview of the PaddlePaddle community, please visit community.
??"PaddleOCR 2.9 released, officially open source text image intelligent analysis tool", text image layout analysis to achieve high-precision real-time prediction, low-code full-process development to accelerate industrial applications. Integrated text image correction, layout area detection, regular text detection, seal text detection, text recognition, form recognition and other functions. Six model production lines can be called with one click, significantly reducing development costs. Supports multiple deployment methods such as high-performance inference, service-based deployment, and client-side deployment. The live broadcast at 19:00 on October 24th (Thursday) will give you an in-depth analysis of the highlights of the latest upgrade. Registration link
?2024.10.1 Add low-code full-process development capabilities in the OCR field :
? Model enrichment with one click call : Integrate 17 models related to text image intelligent analysis, general OCR, general layout analysis, general table recognition, formula recognition, and seal text recognition into 6 model production lines, through a minimalist Python API Key call to quickly experience the model effect. In addition, the same set of APIs also supports a total of 200+ models such as image classification, target detection, image segmentation, and time series prediction, forming 20+ single-function modules to facilitate developers to use model combinations .
? Improving efficiency and lowering the threshold : Provides two methods based on unified commands and graphical interface to realize simple and efficient use, combination and customization of models. Supports multiple deployment methods such as high-performance inference, service-based deployment, and client-side deployment . In addition, you can seamlessly switch between various mainstream hardware such as NVIDIA GPU, Kunlun Core, Ascend, Cambrian, and Haiguang when developing models.
PaddleX, a low-code development tool, relies on PaddleOCR's advanced technology to support low-code full-process development capabilities in the OCR field:
Supports document scene information extraction v3PP-ChatOCRv3-doc, high-precision layout area detection model based on RT-DETR and PicoDet's high-efficiency layout area detection model, high-precision table structure recognition model SLANet_Plus, text image correction model UVDoc, and formula recognition model LatexOCR , Document image orientation classification model based on PP-LCNet
?2024.7 Add the PaddleOCR Algorithm Model Challenge champion solution :
Competition question 1: OCR end-to-end recognition task champion solution - scene text recognition algorithm-SVTRv2;
Competition question 2: The champion solution for the general table recognition task - table recognition algorithm-SLANet-LCNetV2.
It supports a variety of OCR-related cutting-edge algorithms, and on this basis creates industrial-grade characteristic models PP-OCR, PP-Structure and PP-ChatOCR, and connects the entire process of data production, model training, compression, and prediction deployment.
For complete documentation, please go to: docs
"Learning OCR Hands-on" e-book
This project is released under the Apache License Version 2.0.
