Chunkr Layout Bench - Open Source Subset
Open-source subset of the Chunkr Layout Detection benchmark dataset, containing 360 professionally annotated documents with 4,695 layout annotations across 16 element classes.
This dataset benchmarks document layout detection models on diverse, real-world documents including financial reports, legal contracts, research papers, medical records, and more.
It draws from a curated collection of publicly accessible or research datasets listed below (grouped by domain) and augmented with select government and open-document sources.
🧩 Dataset Overview
Document Categories
- Financial (16.7%), Legal (11.1%), Medical (10.3%), Government (8.1%)
- Technical, Research, Historical, Real Estate, Education
- Construction, Procurement, Tax, Billing, Supply Chain
- Textbook, Consulting, Newspaper, Patent, Magazine, Miscellaneous
Layout Diversity
- 46% Single-column documents
- 42% Multi-column documents
- 12% Double-column documents
📊 Element Classes
| Class |
ID |
Count |
Description |
| Text Block |
1 |
1,717 |
Main body text regions |
| List Item |
11 |
769 |
Bulleted or numbered list elements |
| Title |
8 |
488 |
Section and document titles |
| Line Number |
10 |
310 |
Line numbering (code, legal docs) |
| Page Number |
12 |
260 |
Page numbering elements |
| Header |
7 |
217 |
Document headers |
| Footer |
2 |
196 |
Document footers |
| Table |
14 |
195 |
Tabular data structures |
| Form Region |
4 |
190 |
Form fields and structures |
| Picture |
13 |
91 |
Images and photographs |
| Graphical Item |
6 |
82 |
Charts, diagrams, illustrations |
| Footnote |
3 |
76 |
Footnote text |
| Caption |
0 |
52 |
Figure and table captions |
| Formula |
5 |
31 |
Mathematical expressions |
| Legend |
9 |
15 |
Chart and diagram legends |
| Unknown |
15 |
6 |
Unclassified elements |
📁 Dataset Structure
chunkr-layout-bench-oss/
├── images/ # 360 document images (PNG format)
├── _annotations.coco.json # COCO format annotations
├── metadata.csv # Image metadata (doc_category, width, height, id)
└── analysis/ # Statistical analysis and visualizations
├── dataset_overview.png
├── source_raw_layouts.png
├── dataset_summary.png
└── ...
🖼️ Sample Annotations
📈 Dataset Statistics
Layout Examples
⚙️ Usage
Load with Hugging Face datasets
from datasets import load_dataset
dataset = load_dataset("ChunkrAI/chunkr-layout-bench-oss")
for item in dataset['train']:
print(item['doc_category'], item['width'], item['height'])
With Supervision
import supervision as sv
dataset = sv.DetectionDataset.from_coco(
images_directory_path="images",
annotations_path="_annotations.coco.json"
)
print(f"Dataset size: {len(dataset)}")
print(f"Classes: {dataset.classes}")
Convert to YOLO Format
import supervision as sv
dataset = sv.DetectionDataset.from_coco(
images_directory_path="images",
annotations_path="_annotations.coco.json"
)
dataset.as_yolo(
images_directory_path="yolo/images",
annotations_directory_path="yolo/labels",
data_yaml_path="yolo/data.yaml"
)
📚 Data Attribution and References
The following tables summarize datasets and sources referenced for sampling, benchmarking, or cross-domain balancing in Chunkr Layout Bench OSS.
All sources are publicly accessible for research or under permissive licenses (MIT, CC-BY, public domain).
🧾 Financial & Business Documents
⚖️ Legal, Government & Regulatory
🏥 Medical & Healthcare
| Dataset / Source |
Domain |
Access Location |
| Synthetic Medical Forms / OMR Scanned Medical Forms |
Healthcare / Admin |
Hugging Face Datasets |
| Synthea Synthetic Patient Records |
Healthcare (EHR) |
AWS Open Data Registry |
| Blue Cross Blue Shield Sample EOBs |
Healthcare / Insurance |
Public Sample PDFs |
| Handwritten Rx (Kaggle) |
Prescriptions |
Kaggle Dataset |
🏢 Real Estate, Construction & Engineering
📚 Academic, Historical & Educational
🧮 Technical, Manufacturing & Miscellaneous
| Dataset / Source |
Domain |
Access Location |
| PRImA Layout Analysis |
Technical Publications |
PRImA Dataset |
| Safety Data Sheets (SDS) Online Repositories |
Manufacturing / Chemical |
SDSManager / Chemical Safety |
| Disassembly BOM Dataset |
Manufacturing / Electronics |
Figshare |
| Tobacco3482 / Tobacco800 |
Business Records |
Kaggle Dataset |
| Customs Import Declarations (Synthetic) |
Logistics / Customs |
GitHub Dataset |
🧾 License
MIT License
Free to use for research and commercial purposes.
Where applicable, individual third-party datasets retain their original licenses, all of which allow redistribution or benchmarking under fair-use or research terms.
📎 Citation
If you use this dataset or benchmark, please cite:
@dataset{chunkr_layout_bench_oss_2025,
title = {Chunkr Layout Bench - Open Source Subset},
author = {ChunkrAI},
year = {2025},
publisher = {Lumina AI Inc.},
license = {MIT},
url = {https://huggingface.co/datasets/ChunkrAI/chunkr-layout-bench-oss}
}
