alakxender's picture
Update README.md
6f842e5 verified
metadata
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
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.