Description
Extracting tables and paragraphs from scans and PDFs with plain OCR yields unstructured text and broken tables. deepdoctection is a Python document AI framework that chains layout detection, table recognition, OCR and language models into one pipeline with structured output.
Version 1.0 is PyTorch-only, supports fine-tuned models like LayoutLM and LiLT from Hugging Face, and is split into smaller packages.
Layout analysis:Titles, paragraphs, figures and tables.
Table recognition:Rebuilds rows and columns.
OCR integration:Several OCR engines.
Model ecosystem:Document models from Hugging Face.
Version 1.0 is PyTorch-only, supports fine-tuned models like LayoutLM and LiLT from Hugging Face, and is split into smaller packages.
Features
Layout analysis:Titles, paragraphs, figures and tables.
Table recognition:Rebuilds rows and columns.
OCR integration:Several OCR engines.
Model ecosystem:Document models from Hugging Face.
