MyanmarOCR-ImageText: 41,664 Burmese Images with 32 Variations per Text

41,664 images of Burmese text at 512 × 512 resolution comprise this dataset for OCR and multimodal AI research. It contains 1,139 unique Burmese text entries, each rendered in 32 stylistic variations, and was created by author kalixlouiis. The dataset was last updated on 2026-04-23.

Use Cases

• Train optical character recognition models based on the clean and diverse Burmese image-text pairs.
• Fine-tune vision-language models based on the image-to-text data.
• Benchmark multimodal AI systems on a dedicated Burmese-language dataset.
• Research font and style invariance in text recognition based on the 32 variations per text entry.

Strengths

• Contains 41,664 high-resolution images at 512 × 512 pixels.
• Offers 32 stylistic variations for each of the 1,139 unique Burmese text entries.
• Focuses exclusively on Burmese text with no English words or punctuation, providing a clean language-specific resource.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• The dataset is split for training only, which may limit its use for validation or testing without further partitioning.
• Row count is unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Freshness

Last updated 2026-04-23 15:18:15; freshness should be verified.