Tabulus

Tabulus#

Scientific PDF Table Extraction Pipeline

About Tabulus#

Tabulus is a modular pipeline for digitizing scientific PDF papers into structured, reference-aware table data.

Tabulus workflow overview

The rebuilt library is organized around standalone commands and explicit filesystem contracts. The current rebuilt library covers PDF profiling, canonical table-crop export, table reconstruction, reference-table classification, GROBID-backed bibliography extraction, deterministic reference matching, Step 6 paper-level scholarly reference resolution, and Step 7 resolved CSV export. Run reports and complete tabulus run orchestration remain planned in this checkout. Bibliography extraction is a parallel PDF-level branch that produces references/bibliography.json, not a consumer of MinerU table crops or reconstruction prediction CSVs.

Start with the locally verified commands, then use the linked pages for setup and adapter details. For one PDF:

tabulus profile --pdf /path/to/paper.pdf --backend pipeline

tabulus reconstruct-tables \
  --crops /path/to/tabulus-output/table-crops/<paper> \
  --adapter <adapter> \
  --device gpu:0

tabulus classify-reference-tables \
  --reconstruction /path/to/tabulus-output/table-crops/<paper>/reconstructions/<adapter>

tabulus extract-bibliography \
  --pdf /path/to/paper.pdf \
  --out /path/to/artifact-root \
  --grobid-url http://localhost:8070

tabulus match-references \
  --selected /path/to/selected_reference_tables.json \
  --bibliography /path/to/artifact-root/references/bibliography.json

tabulus resolve-references \
  --bibliography /path/to/artifact-root/references/bibliography.json \
  --reference-matches /path/to/reconstruction/references/reference_matches.json \
  --out /path/to/artifact-root

tabulus export-resolved-csv \
  --reference-matches /path/to/reconstruction/references/reference_matches.json \
  --reference-resolution /path/to/artifact-root/references/reference_resolution.json

For several PDFs in one folder:

tabulus profile \
  --folder /path/to/papers \
  --backend hybrid-engine \
  --method auto \
  --effort high

tabulus reconstruct-tables \
  --crops-folder /path/to/papers/tabulus-output/table-crops \
  --adapter <adapter> \
  --device gpu:0

tabulus classify-reference-tables \
  --crops-folder /path/to/papers/tabulus-output/table-crops \
  --adapter <adapter>

How To Use This Documentation#

Start with the installation page that matches your machine. Windows and CPU-only users can use MinerU’s pipeline backend. GPU-server users can use MinerU’s hybrid-engine backend and adapter-specific reconstruction environments when a suitable CUDA GPU is visible. Library contributors can use the core Python setup for GPU-independent unit tests.

For the current adapter list and command examples, see Step 2: Table Reconstruction.

Where To Start#

CPU / Windows Setup

Use Python 3.12, a standard venv, CPU-only PyTorch, and MinerU pipeline.

Windows CPU Setup
GPU Server Setup

Request GPU resources, install step-specific environments, and run MinerU or Step 2 reconstruction adapters.

GPU Server Installation
Install The Python Library

Install Tabulus for core library development and run GPU-independent unit tests.

Python Library Setup
Run MinerU On GPU

Use the tested MinerU 3.4.5 command sequence and validate the output with discover_tables.

MinerU GPU Execution
Follow The Core Tutorial

Read the current runnable steps and artifact flow.

Core Pipeline Overview

Documentation Map#

The sidebar contains the full documentation. The main sections are:

  • Installation And Setup: Windows CPU, GPU server, and Python library setup.

  • Tutorial: the intended modular workflow, one processing step at a time.

  • Components: adapter boundaries and responsibilities.

  • Workflows: headless local/GPU execution shapes.

  • External Tools: third-party tools as used by Tabulus.

  • Data Contracts: the file formats and artifact layers exchanged between modules.

  • Evaluation: how to compare extraction and matching quality.