Mozhi-Malayalam: Printed Word Images and Transcriptions for Malayalam

A filtered collection of printed word images paired with their corresponding transcriptions in the Malayalam language. The dataset is a Malayalam-only subset derived from the 'darknight054/indic-mozhi-ocr' collection and was uploaded by author 'trysem'. The dataset listing was last updated on 2026-06-13.

Use Cases

• Train OCR models for Malayalam script based on the described image-text pairs.
• Benchmark text recognition accuracy for printed Malayalam documents.
• Fine-tune pre-trained multilingual OCR models on a specific Indic language subset.
• Create synthetic data pipelines for augmenting Malayalam text recognition datasets.

Strengths

• Focuses on a single, specific language (Malayalam), which may reduce noise for targeted model training.
• The description explicitly states the dataset contains paired image and transcription data, a core requirement for supervised OCR tasks.

Limitations

• Description metadata is limited; actual data quality, image resolution, and transcription accuracy require manual inspection after download.
• Row count is unknown, which may limit suitability assessment for large-scale model training.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

Filtered version of 'darknight054/indic-mozhi-ocr' dataset.

Collection Method

Likely extracted via OCR from printed Malayalam sources, though the specific method is not detailed.

Freshness

Last updated 2026-06-13 16:02:07; freshness should be verified.