23 tools
Extracts tables from text-based PDFs (not scanned documents) into pandas DataFrames, exportable to CSV/JSON/Excel/HTML/SQLite, with accuracy/whitespace quality metrics and a CLI. Five parsing strategies: `lattice` (ruled tables), `stream` (whitespace-based), `network`/`hybrid` (text-alignment), and an optional neural `ml` parser (Table Transformer) for hard borderless cases.
PythonCog/Replicate-deployable packaging of Marker (see the separate Marker entry), converting PDFs — and EPUB, MOBI, XPS, FB2 — to Markdown via layout analysis and OCR, with table formatting and LaTeX-equation conversion carried over from the underlying Marker engine. Exists as its own entry because it's a distinct deployable artifact (a Cog container image runnable on Replicate's infrastructure) rather than a different tool — the actual conversion logic is Marker's.
Apache-2.0 orchestration framework for document AI: combines layout detection and table recognition (via Detectron2/Transformers), OCR (Tesseract, DocTr, or AWS Textract), and document/token classification (LayoutLM family, LiLT, BERT-style models) into configurable pipelines for scanned documents and PDFs.
PythonDocument-conversion toolkit that parses PDFs (and Office/image formats) into a unified structured representation, with layout analysis, OCR (EasyOCR by default), table structure recognition, and optional vision-language-model image description; exports to Markdown, HTML, or JSON.
PythonConverts PDFs, images, PPTX, DOCX, XLSX, HTML, and EPUB to Markdown, JSON, HTML, or chunks, preserving reading order: reconstructs tables (heuristics with VLM fallback for complex cases), recognizes math/equations as LaTeX, extracts and saves images, performs multilingual OCR via the Surya VLM, strips headers/footers/artifacts, and can optionally use an LLM to improve accuracy.
PythonConverts PDF, DOCX, PPTX, XLSX, images, and web pages into Markdown or JSON: strips headers/footers/page numbers while preserving structure, converts formulas to LaTeX and tables to HTML, OCRs scanned/garbled PDFs in 109 languages, handles single/multi-column and complex layouts (incl. cross-page table merging), and extracts images with captions. Offers CLI, FastAPI, and Gradio WebUI, plus Python/Go/TypeScript SDKs; builds on PDF-Extract-Kit's models.
PythonConverts scanned-book PDFs to Markdown or EPUB using OCR (DeepSeek OCR locally, or DeepSeek/Baidu vendor backends), recognizing complex structure including tables and formulas; extracts body text while filtering headers/footers, preserves footnotes/images/assets, auto-generates an EPUB table of contents (optionally LLM-enhanced), and offers configurable table (HTML/image) and formula (MathML/SVG/image) rendering. Runs fully offline as of v1.0.0.
PythonModel toolbox for extracting content from complex PDFs: layout detection (images/tables/text/titles/formulas), formula detection and LaTeX recognition, OCR with text location, and table recognition to LaTeX/HTML/Markdown. Provides the underlying models that MinerU builds its PDF-to-Markdown application on top of.
JavaScriptCross-platform TypeScript/JavaScript module extracting text, embedded images, tables, and document metadata/outline from PDFs, plus rendering pages to PNG screenshots — a broader capability set than most JS PDF-parsing libraries, which typically stop at plain text. A complete rewrite under new maintainership (mehmet-kozan) since the original gitlab.com/autokent/pdf-parse went unmaintained, actively published through late 2025.
Command-line tool for extracting text and tabular data from PDF files, with table output written directly to CSV for further processing. Built as a focused, single-purpose alternative to heavier PDF libraries for when all that's needed is getting content out of a document, without pulling in a full parsing/rendering engine as a dependency. Written by the same author as the PDF-Writer/HummusJS engine, sharing some of its low-level PDF parsing code.
PythonConverts PDF files to DOCX (Word) format, preserving formatting and layout — fonts, positioning, images — rather than dumping plain text into a Word document, and extracts tables as genuine Word tables rather than flattened text. Offers both a CLI and a GUI, covering scripted batch conversion and one-off interactive use. A common need for taking a PDF that needs further editing back into an editable Word document.
ScalaExtracts figures, tables, captions, figure names/numbers, section titles, and in-figure text from scholarly (computer science) PDFs, with page location and bounding boxes; renders extracted figures as raster (PNG/JPEG) or vector (SVG/PS/EPS, via pdftocairo) images; supports multi-threaded batch processing and includes debugging visualization tools.
Subclass of Apache PDFBox's own `PDFTextStripper` specifically built to preserve a PDF's original layout and spatial positioning when converting it to plain text — where PDFBox's default stripper reflows text into reading order and loses column/table alignment, this one keeps whitespace positioned to match the source layout. Makes it noticeably easier to extract structured data such as tables and form content from the resulting plain text, at the cost of extra leading whitespace in the output.
Extracts text (with layout preservation, tolerance settings, regex search, and word bounding boxes), detects and extracts tables (via ruling-line or text-alignment strategies), and provides low-level access to characters, lines, rectangles, curves, images, annotations, hyperlinks, and form fields, with page cropping/filtering, password-protected PDF support, and visual debugging (page-to-image rendering with overlay annotations, incl. Jupyter display). Does not generate, modify, or OCR PDFs.
Node.js library that reads text from PDFs and parses tabular data using automatic column detection combined with rule-based parsing rather than a fixed grid layout. Aimed specifically at extracting structured data (tables, forms) from PDFs programmatically, rather than general-purpose text dumping — the rule-based parser lets callers define matching patterns for specific document layouts they expect to process repeatedly.
PythonToolkit for extracting tabular data from OCR'd ('sandwich') PDFs already converted to pdf2xml format: splits scanned double pages, detects lines via image processing, corrects page skew/rotation, clusters lines/text into table columns and rows, and exports to pandas DataFrames (CSV, Excel, etc.). Not an OCR tool itself — requires pre-OCR'd input from Tesseract/ABBYY; the authors recommend trying `pdftotext` first.
Pipeline library for extracting information from unstructured documents with low memory/CPU overhead: PDF table extraction, image analysis with OCR, document question-answering via LLM integration, vector index creation, and support for most common document formats.
Python bindings for MuPDF covering text extraction (plain, rich dictionary with font/size/color/bbox, HTML/XML, raw blocks, table-to-Markdown); embedded image extraction and high-DPI page rendering to Pixmaps; Tesseract-based OCR (100+ languages); reading/writing annotations (highlights, notes, stamps, ink) and redactions (with permanent apply); AcroForm reading and filling; page insert/delete/reorder, merge/split, and metadata editing; PDF creation and Markdown-to-PDF conversion; password protection/RC4/AES encryption; hyperlink and cross-reference handling; and outline/bookmark read-write.
PythonMarkdown-conversion layer built on PyMuPDF: converts PDF pages to Markdown while preserving headings, tables (ruled-line or text-position-based detection), and images, specifically tuned for feeding PDF content to LLMs as clean, structured Markdown rather than raw extracted text. A sibling package to base PyMuPDF rather than a separate engine — it reuses PyMuPDF's own C-level parsing and adds the Markdown-shaping layer on top.
Table-extraction library (powers the Tabula tool) that pulls tabular data out of PDFs using lattice mode (ruled cell borders) or stream mode (no borders), exporting to CSV, TSV, or JSON; usable as a CLI for batch processing or as a Java API, with support for selective pages, custom extraction areas, and password-protected files.
R bindings to the Tabula Java library (an R clone of the Tabula CLI) that computationally extracts tables from PDF documents via `extract_tables()`, returning results as R tibbles/data frames. Requires a Java runtime (rJava, OpenJDK 11 recommended).
GoComprehensive Go PDF toolkit for reading, writing, and processing PDFs: text/image/table extraction, merging/splitting/rotating/rearranging pages, form creation and filling, encryption/decryption, digital signing, watermarking, and image-to-PDF conversion.
Go PDF library (pure Go): generates reports/invoices with styled text and tables; creates and fills forms; merges, splits, reorders, and rotates pages (incl. N-up layouts); extracts text (with position/formatting), images, and tabular data to CSV; adds/converts images to PDF; removes or adds password protection; validates and applies digital signatures; compresses/optimizes files; adds watermarks; and supports CCITTFax/JBIG2 image decoding.