ved1245/synthetic-manuscript-dataset

Synthetic Manuscript Dataset

Synthetic historical manuscript folios generated using an automated Python pipeline.

Scripts

The dataset contains three script configurations:

• Devanagari
• Modi
• Sharada

Each script contains 100 synthetic manuscript folios.

Dataset Splits

Split	Samples per Script
Train	85
Validation	10
Test	5
Total	100

Across all three scripts, the dataset contains:

• 300 manuscript images
• 300 corresponding Markdown (.md) annotations

Dataset Organization

devanagari/
├── train/
├── validation/
└── test/

modi/
├── train/
├── validation/
└── test/

sharada/
├── train/
├── validation/
└── test/

Every generated PNG manuscript has a corresponding Markdown ground-truth transcription.

Example:

devanagari_train_001.png
devanagari_train_001.md

Manuscript Styles

The generated folios include:

• Aged handmade paper
• Palm-leaf manuscript backgrounds
• Paper fibres and aging
• Stains and water damage
• Irregular manuscript edges
• Ink fading and smudging
• Calligraphic variation
• Baseline waviness
• Multiple manuscript layouts
• Highlighted manuscript text

Generation

The dataset was produced using the Synthetic Manuscript Generator, an automated Python pipeline supporting Devanagari, Modi, and Sharada scripts.

Each image is accompanied by synchronized ground-truth text in Markdown format.

Author

Vedant Paturkar