Parsynth OCR 200K: A Preview of a 4M-Sample Synthetic Persian OCR Dataset

ParsynthOCR 200K is a synthetic dataset for Persian optical character recognition. It is a preview version of the original 4 million sample dataset, ParsynthOCR-4M. The dataset was uploaded by hezarai and last updated on May 7, 2024.

Use Cases

• Training Persian OCR models based on synthetic text images.
• Benchmarking OCR performance on Persian script.
• Fine-tuning multilingual OCR systems with Persian data.
• Developing synthetic data generation techniques for non-Latin scripts.

Strengths

• Synthetic nature allows for controlled generation of text-image pairs.
• Preview of a larger 4 million sample dataset, indicating potential scale.
• Specifically designed for Persian OCR, addressing a niche language need.

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 is unknown, which may limit suitability assessment.

Provenance

Source

hezarai

Collection Method

Synthetically generated.

Time Range

null

Freshness

Last updated 2024-05-07 08:55:33; freshness should be verified.

Geography

null