ignaciaginting's picture
Upload 396 files
230c9a6 verified

A newer version of the Streamlit SDK is available: 1.45.1

Upgrade

PDF2Markdown

Demo:(left: input image; right: rendered markdown.)

demo

  1. Extract PDF features by these tasks:

    • Layout Detection: Using the YOLOv8 model for region detection, such as images, tables, titles, text, etc.;

    • Formula Detection: Using YOLOv8 for detecting formulas, including inline formulas and isolated formulas;

    • Formula Recognition: Using UniMERNet for formula recognition;

    • Table Recognition: Using StructEqTable for table recognition;

    • Optical Character Recognition: Using PaddleOCR for text recognition;

  2. Convert features to markdown file:

    Using simple rules to convert the identified result to markdown (Note: this is a simply convert code and can only support one-column PDFs, see MinerU for more complex situation).

Usage

python project/pdf2markdown/scripts/run_project.py --config project/pdf2markdown/configs/pdf2markdown.yaml