Dataset Viewer
The dataset viewer is not available for this dataset.
Job has been terminated due to a temporary spike in resource usage and may be restarted later.
Error code:   JobManagerCrashedError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

aCROne

aCROne is a synthetic OCR dataset generated with a Python-based image generator for training and evaluating OCR models.

Dataset

The dataset contains synthetic images of:

  • words
  • sentences
  • random characters

Samples are generated using multiple visual styles:

  • clean
  • printed
  • handwritten
  • distorted
  • cursed

Images can contain different fonts, backgrounds, noise, rotations, scaling, brightness/contrast changes, and CAPTCHA-like distortions.

The dataset contains approximately 500,000 images.

Data format

Each sample contains an image and its corresponding text transcription.

The dataset is split into:

  • train.json
  • val.json

Each record contains:

{
  "image": "example_0000001.webp",
  "style": "clean",
  "text": "example"
}

Intended use

aCROne is intended for:

  • OCR model training
  • OCR evaluation
  • computer vision experiments
  • synthetic data generation research

License

MIT

Downloads last month
296