OCR Data 22K: A Subset for Training Optical Character Recognition Models

A subset of the TextOCR split from Yesiarohn/OCR-Data, created by theminji on 2026-05-15. This dataset is intended for training and testing optical character recognition models, specifically designed to be a smaller, more focused collection for experimentation.

Use Cases

• Fine-tuning OCR models based on the described text-in-image data.
• Benchmarking model performance on a focused subset of OCR tasks.
• Testing vision-language models on a smaller, curated dataset for faster iteration.

Strengths

• Derived from the established TextOCR split, suggesting a known source structure.
• Explicitly curated to be a smaller, more manageable dataset for testing purposes.

Limitations

• Description metadata is limited; actual data quality requires manual inspection after download.
• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and file formats are unknown, which may limit suitability assessment.

Provenance

Source

Yesiarohn/OCR-Data (TextOCR split)

Collection Method

Subset created by the author for testing purposes.

Freshness

Last updated 2026-05-15 06:35:52; freshness should be verified.