Showcase/OmniDocBench Comprehensive Document Parsing

advancedtext

OmniDocBench Comprehensive Document Parsing

Comprehensive document parsing annotation covering layout detection, text recognition, table structure, and formula recognition. Annotators draw bounding boxes and provide text transcriptions for document elements.

Configuration Fileconfig.yaml

This Potato config reproduces the annotation task. Save it as config.yaml and run potato start config.yaml to try it.

yaml

# OmniDocBench Comprehensive Document Parsing Configuration
# Based on Chen et al., CVPR 2025

annotation_task_name: "OmniDocBench Document Parsing"
task_dir: "."

data_files:
  - "sample-data.json"

item_properties:
  id_key: "id"
  text_key: "image_url"
  context_key: "document_type"

user_config:
  allow_all_users: true

annotation_schemes:
  - annotation_type: "text"
    name: "element_bounding_boxes"
    description: "Draw bounding boxes around each document element (format: element_type,x,y,width,height per line). Element types: Text, Title, Table, Figure, Formula, Header, Footer, Caption, Reference, Abstract"

  - annotation_type: "text"
    name: "ocr_transcription"
    description: "Provide OCR text transcription for each detected element (format: element_id|transcribed_text per line)"

  - annotation_type: "text"
    name: "table_structure"
    description: "For each table element, describe its structure (format: table_id|num_rows,num_cols|cell_content_summary)"

  - annotation_type: "text"
    name: "formula_latex"
    description: "For each formula element, provide the LaTeX transcription (format: formula_id|latex_expression)"

  - annotation_type: "multiselect"
    name: "element_types_present"
    description: "Select all element types present on this document page"
    labels:
      - name: "Text"
        tooltip: "Regular paragraph body text"
      - name: "Title"
        tooltip: "Document title or major headings"
      - name: "Table"
        tooltip: "Tabular data with rows and columns"
      - name: "Figure"
        tooltip: "Images, charts, diagrams, plots"
      - name: "Formula"
        tooltip: "Mathematical equations or expressions"
      - name: "Header"
        tooltip: "Page headers or running titles"
      - name: "Footer"
        tooltip: "Page footers, page numbers"
      - name: "Caption"
        tooltip: "Captions for figures or tables"
      - name: "Reference"
        tooltip: "Bibliography or reference entries"
      - name: "Abstract"
        tooltip: "Paper or document abstract section"

  - annotation_type: "radio"
    name: "document_language"
    description: "What is the primary language of this document?"
    labels:
      - name: "english"
        tooltip: "English language document"
      - name: "chinese"
        tooltip: "Chinese language document"
      - name: "mixed"
        tooltip: "Multiple languages present"
      - name: "other"
        tooltip: "Other language"

  - annotation_type: "radio"
    name: "parsing_difficulty"
    description: "Rate the overall parsing difficulty of this document page"
    labels:
      - name: "easy"
        tooltip: "Simple layout, clear text, no complex elements"
      - name: "moderate"
        tooltip: "Some tables or figures, standard formatting"
      - name: "hard"
        tooltip: "Complex tables, formulas, multi-column, or degraded quality"
      - name: "very_hard"
        tooltip: "Highly complex layout with many overlapping element types"

interface_config:
  item_display_format: "<img src='{{text}}' style='max-width:100%; max-height:600px; border:1px solid #ccc;'/><br/><small>Document type: {{document_type}} | Page: {{page_number}} | Source: {{source}}</small>"

output_annotation_format: "json"
output_annotation_dir: "annotations"

Sample Datasample-data.json

json

[
  {
    "id": "omnidoc_001",
    "image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/8/8e/Pubmed_central_abstract.png/800px-Pubmed_central_abstract.png",
    "document_type": "scientific_paper",
    "page_number": 1,
    "source": "arxiv"
  },
  {
    "id": "omnidoc_002",
    "image_url": "https://upload.wikimedia.org/wikipedia/commons/thumb/2/2e/US_patent_1.png/800px-US_patent_1.png",
    "document_type": "patent",
    "page_number": 1,
    "source": "uspto"
  }
]

// ... and 8 more items

Try it live — no install

Boot the real Potato server in your browser (WebAssembly) and annotate with this exact config. Nothing leaves your machine.

▶ Run live in your browser

Get This Design

View on GitHub

Clone or download from the repository

Quick start:

git clone https://github.com/davidjurgens/potato-showcase.git
cd potato-showcase/image/omnidocbench-document-parsing
potato start config.yaml

Dataset & paper

Ouyang et al., CVPR 2025

Official dataset ↗Read the paper ↗

Citation (BibTeX)

bibtex

@inproceedings{ouyang2025omnidocbench,
    title = "{OmniDocBench}: Benchmarking Diverse {PDF} Document Parsing with Comprehensive Annotations",
    author = "Ouyang, Linke and Qu, Yuan and Zhou, Hongbin and Zhu, Jiawei and Zhang, Rui and Lin, Qunshu and Wang, Bin and Zhao, Zhiyuan and Jiang, Man and Zhao, Xiaomeng and Shi, Jin and Wu, Fan and Chu, Pei and Liu, Minghao and Li, Zhenxiang and Xu, Chao and Zhang, Bo and Shi, Botian and Tu, Zhongying and He, Conghui",
    booktitle = "Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)",
    year = "2025",
    url = "https://arxiv.org/abs/2412.07626"
}

Details

Annotation Types

multiselectradiotext

Domain

Document AIPDF Parsing

Use Cases

Document ParsingLayout DetectionOCRTable RecognitionFormula Recognition

Related Designs

DocBank Document Layout Detection

Document layout analysis benchmark (Li et al., COLING 2020). Detect and classify document elements including titles, abstracts, paragraphs, figures, tables, and captions.

multiselectradio

DocLayNet Document Layout Analysis

Document layout analysis with bounding box annotations. Annotators draw bounding boxes around layout elements (text blocks, tables, figures, headers, footers, lists) in document page images.