jonny122/khmer-newspaper-layout-dataset

Khmer Newspaper Layout Dataset

Dataset Description

This dataset contains Khmer newspaper layouts with annotated regions for document layout analysis and OCR tasks.

Dataset Summary

• Total Examples: 9,344 newspaper layouts
• Language: Khmer (Cambodian)
• Image Format: PNG
• Annotations: LabelMe JSON format with bounding boxes and segmentation masks

Features

• config_id: Unique identifier for each sample
• image: Newspaper layout image (PNG)
• article_title: Article title in Khmer
• article_subtitle: Article subtitle
• article_body: Full article text in Khmer
• article_author: Author name
• article_date: Publication date
• article_section: Newspaper section
• article_location: Location information
• article_edition: Edition information
• column_count: Number of columns (1 or 2)
• style: Layout style (e.g., magazine, elegant, vintage, classic, minimal, dark)
• font: Khmer font used
• paper_effects: JSON string with paper texture and effects
• labelme_annotations: JSON string with bounding boxes and segmentation masks
• num_annotations: Number of annotated regions

Use Cases

• Document layout analysis
• Khmer OCR training and evaluation
• Document understanding models
• Text detection and recognition
• Page segmentation

Dataset Statistics

• Images: 9,344
• Average annotations per image: ~150-200 regions
• Column layouts: 1-column and 2-column formats
• Styles: Multiple newspaper styles (magazine, elegant, vintage, etc.)

Data Format

Each example includes:

1. Image: Full newspaper layout
2. Text Content: Article title, subtitle, body, and metadata
3. Annotations: Bounding boxes and masks for text regions, images, headers, etc.

Loading the Dataset

from datasets import load_dataset

# Load the dataset
dataset = load_dataset("jonny122/khmer-newspaper-layout-dataset")

# Access a sample
sample = dataset['train'][0]
print(f"Config ID: {sample['config_id']}")
print(f"Image shape: {sample['image'].size}")
print(f"Number of annotations: {sample['num_annotations']}")

Citation

If you use this dataset, please cite:

@dataset{khmer_newspaper_layout_2025,
  title={Khmer Newspaper Layout Dataset},
  author={Sengngoun},
  year={2025}
}

License

This dataset is released under CC-BY-4.0 license.