ghimireplan/AI

Devanagari Character Recognition

This model recognizes Devanagari characters (numerals, vowels, and consonants) from images. It uses a Convolutional Neural Network (CNN) architecture trained on the NHCD (Nepali Handwritten Character Dataset).

Model Architecture

The model uses a CNN architecture with the following components:

• Multiple Conv2D layers with BatchNormalization
• MaxPooling2D layers for dimensionality reduction
• Dropout layers for regularization
• Dense layers for classification

Usage

You can use this model in two ways:

1. Using the Gradio Interface:

   • Run python app.py
   • Open the provided local URL in your browser
   • Upload an image of a Devanagari character
   • Get the prediction results

2. Using the Model Directly:

from tensorflow.keras.models import load_model
import cv2
import numpy as np

# Load the model
model = load_model('models/devanagari_model.h5')

# Load and preprocess image
image = cv2.imread('path_to_image.jpg', cv2.IMREAD_GRAYSCALE)
image = cv2.resize(image, (32, 32))
image = image.reshape(1, 32, 32, 1) / 255.0

# Make prediction
prediction = model.predict(image)

Dataset

The model is trained on the NHCD dataset, which contains:

• Numerals (0-9)
• Vowels
• Consonants

Performance

The model achieves:

• Training accuracy: ~97%
• Validation accuracy: ~95%
• Test accuracy: ~95%

Requirements

Install the required packages:

pip install -r requirements.txt

License

This project is licensed under the MIT License - see the LICENSE file for details.