iapp/thai_handwriting_dataset

---
license: apache-2.0
task_categories:
- text-to-image
- image-to-text
language:
- th
tags:
- handwriting-recognition
- ocr
pretty_name: Thai Handwriting Dataset
size_categories:
- 10K<n<100K
maintainer: Kobkrit Viriyayudhakorn ([e-mail])
dataset_info:
features:
- name: image
dtype: image
- name: text
dtype: string
- name: label_file
dtype: string
---

# Thai Handwriting Dataset

This dataset combines two major Thai handwriting datasets:

1. BEST 2019 Thai Handwriting Recognition dataset (train-0000.parquet)
2. Thai Handwritten Free Dataset by Wang (train-0001.parquet onwards)

## Maintainer
[e-mail]

## Dataset Description

### BEST 2019 Dataset
Contains handwritten Thai text images along with their ground truth transcriptions. The images have been processed and standardized for machine learning tasks.

### Wang Dataset
- Exclusively focuses on handwritten sentences in Thai language
- Contains 4,920 unique sentences covering various topics and themes
- Created from contributions by 2,026 users, ensuring diverse handwriting styles
- Encompasses various linguistic patterns, vocabulary, and sentence structures

## Dataset Structure

The dataset is provided in parquet file format with the following columns:

- `image`: Image data (Image type)
- `text`: Ground truth transcription of the handwritten text (String)
- `label_file`: Source label file name (String)

## Usage

This dataset is ideal for:
- Handwriting recognition
- Optical character recognition (OCR)
- Natural language processing (NLP)
- Language generation

Researchers, developers, and enthusiasts can utilize this dataset to:
- Develop and benchmark algorithms
- Train machine learning models
- Explore innovative techniques in Thai language analysis and handwriting recognition

## Original Datasets

1. BEST 2019 Thai Handwriting Recognition competition:
https://thailang.nectec.or.th/best/best2019-handwrittenrecognition-trainingset/

2. Thai Handwritten Free Dataset by Wang: Data Market
https://www.wang.in.th/dataset/64abb3e951752d79380663c2