PDF-Extract-Kit/project/pdf2markdown/README.md

PDF2Markdown

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

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