---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: question
    dtype: string
  - name: answer
    dtype: string
  splits:
  - name: train
    num_bytes: 6326810831.2
    num_examples: 379544
  - name: validation
    num_bytes: 1581702707.8
    num_examples: 94886
  download_size: 7863433048
  dataset_size: 7908513539.0
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
task_categories:
- image-to-text
- visual-question-answering
language:
- dv
tags:
- thaana
- dhivehi
- ocr
pretty_name: ocr_thaa
license: apache-2.0
---

# Dhivehi Single-Line Text-Image Dataset

A collection of synthetic Dhivehi text images for training and evaluating text-image / vision models etc. 

Each image contains a single line of Dhivehi text with various visual styles and augmentations (Check the config field for more info on the row).

***Note: This dataset is a subset from alakxender/dhivehi-vrd-images.***