Skip to content

Dataset · SWHL

Text recognition test dataset for 8 scenarios

SWHL/text_rec_test_dataset

A small test dataset with Chinese and English text images for evaluating the quality of OCR models.

What's inside

The dataset contains image-text pairs for testing optical character recognition (OCR) algorithms in Chinese and English. Data is gathered from open web sources and divided into 8 scenarios: vertical text (14 examples), long text (18), individual characters/symbols (115), CAPTCHAs, scene text, bank cards, handwriting, and license plates. Total sample size is under 1,000 records.

What you can do with it

  • Benchmarking and validating OCR models on complex text types (CAPTCHAs, handwriting, vertical writing)
  • Quickly assessing recognition accuracy using the TextRecMetric library
Languages
Chinese, English
Size
up to 1K rows
Downloads this month
658
Likes
3

How to load it

from datasets import load_dataset

ds = load_dataset("SWHL/text_rec_test_dataset")