README.md

metadata

license: apache-2.0
language:
  - ru
  - en
  - es
  - fr
  - de
  - it
  - pt
  - zh
  - ja
  - ko
task_categories:
  - image-to-text
  - text-to-image
task_ids:
  - image-captioning
tags:
  - synthetic
  - text
  - multilingual
  - image-captioning
  - computer-vision
  - ocr
  - text-recognition
  - generated
size_categories:
  - 100K<n<1M
pretty_name: TextSynth-100k

TextSynth-100k

A synthetic dataset for training image captioning and text recognition (OCR) models. 100,000 images with text in 10 languages, in different colors, fonts, and on different backgrounds.

What's inside

• 100,000 images sized 224×224, JPG format.
• 200 unique texts in 10 languages.
• 30 text colors and 12 background colors.
• Various Windows fonts: Arial, Times New Roman, Courier New, Calibri, Cambria, Georgia, and others.
• Different font sizes — from 28 to 90 pt.
• Text rotations: 0°, ±15°, ±30°.
• Captions in the format: the text <text> is written in <color> on a <background> background.

Languages

Language	Example
Russian	привет, мир, кот, собака, дом
English	hello, world, cat, dog, house
Spanish	hola, mundo, gato, perro, casa
French	bonjour, monde, chat, chien, maison
German	hallo, welt, katze, hund, haus
Italian	ciao, mondo, gatto, cane, casa
Portuguese	olá, mundo, gato, cachorro, casa
Chinese	你好, 世界, 猫, 狗, 房子
Japanese	こんにちは, 世界, 猫, 犬, 家
Korean	안녕, 세계, 고양이, 개, 집

Structure