# bytedance/Dolphin

The official repo for “Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting”, ACL, 2025.

Repository: https://github.com/bytedance/Dolphin
Canonical: https://ross.abutalabs.com/products/bytedance-dolphin
Language: Python
License: NOASSERTION
License Family: other
Topics: document-analysis, layout-analysis, ocr, parser, pdf, pdf-converter, pdf-parser, python, vlm-ocr
Last push: 2026-03-25T06:34:12+00:00

## Health v2 (maintenance only)
Score: 52/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 74, release rhythm 35, longevity 34
- inputs: {"age_days": 477, "days_push": 161, "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 9049, forks 776 (observed 2026-08-28T04:10:28.144010+00:00)

## What it is
Dolphin is ByteDance's open-source document image parsing model that converts document images and PDFs into structured content using a two-stage layout-analysis and element-parsing architecture. It handles text, tables, formulas, figures, and code blocks in both digital-born and photographed documents.

## Use cases
- parse scanned pdf documents into structured text
- extract tables and formulas from document images
- convert photographed documents to markdown
- analyze document layout and reading order
- ocr for complex document pages
- batch convert pdfs to structured data

## When to choose
- you need structured parsing of mixed document elements like tables, formulas, and figures
- you want a lightweight model with parallel element-wise parsing for efficiency
- you need to handle both digital and photographed documents

## When to avoid
- you only need plain OCR text extraction without layout structure
- you cannot run GPU inference for the vision-language model
- you need a fully permissive license (license is not standard MIT-verified)

## Facets
- artifact type: library
- maturity: active
- function: ocr, parser, pdf, machine-learning, computer-vision, nlp
- domain: pdf, computer-vision, artificial-intelligence, files
- platform: python, cross-platform
- tags: document-parsing, layout-analysis, vision-language-model, pdf-converter, table-recognition, formula-recognition, natural-language-processing, gpu

## Member repositories
- bytedance/Dolphin (main) score 52

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:10:28.144010+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 2026-08-29T17:24:03.021308+00:00, confidence not recorded.
  - readme: https://github.com/bytedance/Dolphin (fetched 2026-08-28T04:10:28.144010+00:00, sha 3827ac3c7e68)
- Data as of 2026-08-30T08:39:29.467469+00:00.