PaddleOCR

PaddleOCR is a toolkit for text recognition and document parsing, installed as a Python package and running on CPU or GPU. Beyond recognising text lines in more than a hundred languages, its PP-StructureV3 component detects page layout, headings and tables, and turns PDF files and images into Markdown or JSON output with coordinates down to table-cell level. There is also a compact vision-language model for parsing complex documents. Engineers use it to pull value tables out of scanned data sheets and standards, or as the engine inside an automation script. Limitation: there is no graphical interface, installation needs Python know-how, and it is much faster with a GPU.