Amharic (Geʽez) Handwritten Character Images, 13,000+ Samples

Amharic (Geʽez script) handwritten character images for OCR and handwriting recognition research. The dataset contains over 13,000 grayscale images across 287 distinct character classes, each sized 32 by 32 pixels. It was created by Yaredoffice and last updated on January 4, 2026.

Use Cases

• Train character-level optical character recognition (OCR) models based on the described handwritten character images.
• Develop handwriting recognition systems for the Geʽez script based on the 287 distinct character classes.
• Benchmark machine learning models on a balanced dataset of over 13,000 grayscale images.
• Conduct research on low-resolution (32x32 pixel) image classification for handwritten scripts.

Strengths

• Contains over 13,000 individual image samples.
• Covers 287 distinct Amharic (Geʽez) character classes.
• Images are consistently sized at 32 by 32 pixels in grayscale format.
• The dataset is described as balanced across all character classes.

Limitations

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

Provenance

Source

Yaredoffice on Hugging Face

Collection Method

Likely collected as handwritten samples for character recognition.

Freshness

Last updated 2026-01-04 03:41:41; freshness should be verified.