mangalathkedar/malayalam_characters

Malayalam Character Dataset

This dataset contains 136,344 images of Malayalam characters, covering 662 distinct classes.

Dataset Structure

The dataset is organized as a single split (train) with the following columns:

• image_path: The character image (128x128 grayscale).
• text: The Malayalam character/word represented in the image.
• label_id: Integer label ID (0-661).
• filename: Original filename.

Classes

The dataset includes:

• Basic consonants and vowels
• Consonants with vowel signs (matras)
• Conjunct characters (koottaksharam)
• Numerals
• Miscellaneous signs

Creation

• Source: Aggregated from multiple sources and augmented.
• Augmentation: Rotation, translation, noise, and scaling applied to expand smaller classes (e.g., vowel signs).
• Format: Uploaded as Parquet for efficiency.

License

[Add License Here - e.g., MIT, CC-BY-4.0]