Synthetic Cyrillic OCR Image Collection with 300,000 Samples

A synthetic dataset of 300,000 images designed for optical character recognition tasks. It contains printed Cyrillic characters including the full Russian alphabet and common punctuation symbols. The dataset was created by author DonkeySmall and was last updated in July 2024.

Use Cases

• Train a vision model to recognize the full Russian Cyrillic alphabet from synthetic image data.
• Benchmark OCR systems on the punctuation symbols and special characters included in the dataset.
• Generate synthetic text for data augmentation in document analysis pipelines focusing on Cyrillic scripts.

Strengths

• Contains 300,000 images, providing a substantial volume for model training.
• Covers the complete Russian Cyrillic alphabet, including both uppercase and lowercase letters.
• Includes a variety of punctuation and special characters for comprehensive text recognition.

Limitations

• All data is synthetically generated, which may not perfectly match real-world document variations.
• The dataset's specific image characteristics, resolution, and noise patterns are not described.
• No information is provided on potential class balance or distribution of characters across the 300,000 images.

Provenance

Source

huggingface

Collection Method

Synthetically generated data.

Freshness

Last updated on 2024-07-19.