doc-analysis/TableBank resource

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1084 stars · 146 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

TableBank is a large image-based benchmark dataset for table detection and table structure recognition, containing 417K high-quality labeled tables extracted from Word and LaTeX documents via weak supervision. It includes trained detection and recognition models built with Detectron2 and OpenNMT, and is distributed via HuggingFace.

Use cases

• train a model to detect tables in document images
• benchmark table structure recognition models
• extract tables from PDFs and research papers
• get labeled training data for document layout analysis
• evaluate table extraction accuracy on a standard dataset
• fine-tune object detection models for table regions

When to choose

• you need large-scale labeled data for table detection or recognition research
• you want a standard benchmark to compare table extraction models
• you are building document layout analysis pipelines involving tables

When to avoid

• you need production-ready table extraction software rather than a research dataset
• your use case is commercial, since the data is research-only
• you need human-annotated rather than weakly supervised labels

Facets

dataset · maturity maintenance

machine-learning computer-vision ocr data-science machine-learning computer-vision pdf python cross-platform table-detection table-structure-recognition benchmark-dataset weak-supervision document-layout-analysis computer-vision-dataset natural-language-processing

Member repositories

For agents

markdown · JSON · MCP: product_card(name="doc-analysis/TableBank")