Turkish OCR Synthetic Dataset with 1,000 Realistic Images

1,000 synthetic images of Turkish text designed for Optical Character Recognition model training. The dataset, created by emredeveloper and last updated in November 2025, contains a mix of handwritten and printed text with realistic augmentations. It fully supports the Turkish alphabet, including characters like Ç, ğ, İ, and Ş.

Use Cases

• Train Turkish OCR models based on the realistic synthetic images.
• Benchmark OCR model performance on handwritten text based on the 70% handwritten content.
• Test OCR model robustness to noise and distortion based on described augmentations like motion blur and perspective distortion.
• Develop models for text extraction from various paper backgrounds based on features like lined, grid, and textured paper.

Strengths

• Contains 1,000 images, providing a substantial base for model training.
• Explicitly supports the full Turkish character set, including Ç, ğ, İ, ı, Ö, ö, Ş, ş, Ü, and ü.
• Specifies a controlled mix of text types: 70% handwritten and 30% printed.
• Describes realistic augmentations like motion blur, noise, and perspective distortion.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data is synthetic, which may not fully capture the complexity of real-world document scans.

Provenance

Source

huggingface

Collection Method

Synthetically generated.

Freshness

Last updated 2025-11-04 17:23:31; freshness should be verified.