# ibm-aur-nlp/PubLayNet

Repository: https://github.com/ibm-aur-nlp/PubLayNet
Canonical: https://ross.abutalabs.com/products/publaynet
Language: Jupyter Notebook
License: NOASSERTION
License Family: other
Last push: 2025-07-09T01:13:59+00:00

## Health v2 (maintenance only)
Score: 46/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 30, release rhythm 35, longevity 100
- inputs: {"age_days": 2680, "days_push": 421, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1060, forks 167 (observed 2026-08-28T04:03:25.834551+00:00)

## What it is
PubLayNet is a large annotated dataset of over 360k document images from PubMed Central, with bounding boxes and polygonal segmentations for layout elements like text, titles, lists, tables, and figures. The repository also hosts pre-trained Faster-RCNN and Mask-RCNN models and ICDAR 2021 Scientific Literature Parsing competition materials.

## Use cases
- train a document layout detection model
- segment text, tables, and figures in scientific paper images
- benchmark object detection models on document images
- extract structure from PDF pages of research articles
- download pre-trained Mask-RCNN for document analysis
- participate in scientific literature parsing competitions

## When to choose
- you need large-scale labeled data for document layout analysis
- you are training or evaluating object detection models on scientific documents
- you want pre-trained models for parsing PubMed-style paper layouts

## When to avoid
- you need table structure recognition rather than layout detection (see PubTabNet)
- you need annotated documents outside the scientific/medical domain
- you need a ready-made application rather than a dataset and models

## Facets
- artifact type: dataset
- maturity: maintenance
- function: computer-vision, ocr, machine-learning, data-science
- domain: computer-vision, machine-learning
- platform: python, cross-platform
- tags: document-layout-analysis, object-detection, image-segmentation, scientific-documents, pubmed, faster-rcnn, mask-rcnn, icdar, natural-language-processing, datasets

## Member repositories
- ibm-aur-nlp/PubLayNet (main) score 46

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:03:25.834551+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T06:56:56.697436+00:00, confidence not recorded.
  - readme: https://github.com/ibm-aur-nlp/PubLayNet (fetched 2026-08-28T04:03:25.834551+00:00, sha 58969915b228)
- Data as of 2026-08-30T08:39:29.467469+00:00.