YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

Devanagari Handwritten OCR Model

Model Summary

This model recognizes handwritten Devanagari characters and digits using a CNN trained on the UCI Devanagari dataset. It supports 46 classes (36 characters + 10 digits).

Intended Use

This model is intended for academic, research, and educational applications for digitizing handwritten Hindi text. Not suitable for production OCR without further validation.

Training Data

  • Dataset: Devanagari Handwritten Character Dataset (UCI)
  • Total Images: ~92,000
  • Image Size: 32x32, grayscale

Model Architecture

  • Input: 32x32 grayscale images
  • Layers: Conv2D → MaxPooling → Dropout → Flatten → Dense → Softmax
  • Optimizer: Adam
  • Loss: CategoricalCrossentropy

Evaluation Metrics

  • Training Accuracy: 97%
  • Validation Accuracy: 91%
  • Loss values plotted for 20 epochs

Limitations

  • Trained only on clean, centered characters
  • Doesn’t handle multi-character words or real-world handwriting noise

License

MIT License

Citation

Harish Phad (2025), Devanagari OCR using CNN, Project-Based Learning 2

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support