Instructions to use PaddlePaddle/PP-DocLayoutV3 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- PaddleOCR
How to use PaddlePaddle/PP-DocLayoutV3 with PaddleOCR:
# 1. See https://www.paddlepaddle.org.cn/en/install to install paddlepaddle # 2. pip install paddleocr from paddleocr import LayoutDetection model = LayoutDetection(model_name="PP-DocLayoutV3") output = model.predict(input="path/to/image.png", batch_size=1) for res in output: res.print() res.save_to_img(save_path="./output/") res.save_to_json(save_path="./output/res.json") - Notebooks
- Google Colab
- Kaggle
What is the input resolution?
#3
by wamreyaz - opened
I see that inference.yml resizes everything to 800x800 without preserving aspect ratio.
Is this how the model was trained? The paper itself provides quite sparse details on how the model was trained and the docs are not that helpful either.
I say this because the model is confusing the reading order in large (8k x 5k) newspapers for example.

