aoiandroid/PP-DocLayoutV3
04
1---2license: apache-2.03pipeline_tag: image-segmentation4tags:5- PaddleOCR6- PaddlePaddle7- image-segmentation8- ocr9- layout10- layout_detection11language:12- en13- zh14- multilingual15library_name: PaddleOCR16---17 18<div align="center">19 20 21<h1 align="center">22 23Layout Analysis Module of PaddleOCR-VL-1.524 25</h1>26 27[](https://github.com/PaddlePaddle/PaddleOCR)28[](https://huggingface.co/PaddlePaddle/PP-DocLayoutV3)29[](https://modelscope.cn/models/PaddlePaddle/PP-DocLayoutV3)30[](https://huggingface.co/spaces/PaddlePaddle/PaddleOCR-VL-1.5_Online_Demo)31[](https://modelscope.cn/studios/PaddlePaddle/PaddleOCR-VL-1.5_Online_Demo/summary)32[](https://discord.gg/JPmZXDsEEK)33[](https://x.com/PaddlePaddle)34[](./LICENSE)35 36**🔥 [Official Website](https://www.paddleocr.com)** | 37**📝 [Technical Report](https://arxiv.org/abs/2601.21957)**38 39</div>40 41 42 43 44## Introduction45 46This is the PP-Doclayoutv3 model weights for the PaddlePaddle framework. Get safetensors weights at [PP-DocLayoutV3_safetensors](https://huggingface.co/PaddlePaddle/PP-DocLayoutV3_safetensors)47 48**PP-DocLayoutV3 is specifically engineered to handle non-planar document images. It can directly predict multi-point bounding boxes for layout elements—as opposed to standard two-point boxes—and determine logical reading orders for skewed and curved surfaces within a single forward pass, significantly reducing cascading errors.** This model is an essential component of PaddleOCR-VL-1.5, providing crucial layout analysis for the high-precision parsing of various real-world documents in PaddleOCR-VL.49 50 51### **Model Architecture** 52 53<div align="center">54<img src="https://raw.githubusercontent.com/cuicheng01/PaddleX_doc_images/refs/heads/main/images/paddleocr_vl_1_5/PP-DocLayoutV3.png" width="800"/>55</div>56 57 58## Visualization59 60 61### Light Variation62 63<div align="center">64<img src="https://raw.githubusercontent.com/cuicheng01/PaddleX_doc_images/refs/heads/main/images/paddleocr_vl_1_5/layout_lighting.jpg" width="800"/>65</div>66 67 68### Skewing69 70<div align="center">71<img src="https://raw.githubusercontent.com/cuicheng01/PaddleX_doc_images/refs/heads/main/images/paddleocr_vl_1_5/layout_skew.jpg" width="800"/>72</div>73 74 75### Screen-photo76 77<div align="center">78<img src="https://raw.githubusercontent.com/cuicheng01/PaddleX_doc_images/refs/heads/main/images/paddleocr_vl_1_5/layout_screen.jpg" width="800"/>79</div>80 81 82### Curving83 84<div align="center">85<img src="https://raw.githubusercontent.com/cuicheng01/PaddleX_doc_images/refs/heads/main/images/paddleocr_vl_1_5/layout_curv.jpg" width="800"/>86</div>87 88 89## Citation90 91If you find PP-DocLayoutV3 helpful, feel free to give us a star and citation.92 93```bibtex94@misc{cui2026paddleocrvl15multitask09bvlm,95 title={PaddleOCR-VL-1.5: Towards a Multi-Task 0.9B VLM for Robust In-the-Wild Document Parsing}, 96 author={Cheng Cui and Ting Sun and Suyin Liang and Tingquan Gao and Zelun Zhang and Jiaxuan Liu and Xueqing Wang and Changda Zhou and Hongen Liu and Manhui Lin and Yue Zhang and Yubo Zhang and Yi Liu and Dianhai Yu and Yanjun Ma},97 year={2026},98 eprint={2601.21957},99 archivePrefix={arXiv},100 primaryClass={cs.CV},101 url={https://arxiv.org/abs/2601.21957}, 102}103```