ronaksp/Gujarati-Digit-Recognizer

✍️ Multi-Digit Gujarati Recognizer

This is a custom-built web application that recognizes handwritten Gujarati digits. It was originally built as a Tkinter desktop application and has been fully ported to the web using Flask and a custom HTML/CSS/JS frontend to perfectly preserve the original dark-themed interface.

🚀 Features

• —Real-time Drawing Canvas: Smooth drawing experience with brush and eraser tools.
• —Multi-Digit Recognition: Uses OpenCV to detect, crop, and isolate individual digits from left to right.
• —High Accuracy Model: Powered by an optimized Convolutional Neural Network (CNN) built with TensorFlow/Keras.
• —Custom UI: Exact replica of the original Desktop Tkinter application.

🛠️ Tech Stack

• —Frontend: HTML5 Canvas, Vanilla JS, CSS3
• —Backend: Python, Flask, OpenCV
• —Machine Learning: TensorFlow, Keras (.h5 model)

📸 How it Works

1. 1.Draw Gujarati digits on the black canvas.
2. 2.The frontend sends the base64 image to the Flask backend.
3. 3.OpenCV processes the image (Thresholding, Dilation, Contour detection).
4. 4.Digits are extracted, resized to 28x28 (MNIST style), and fed into the CNN.
5. 5.The model predicts the digits and returns the result to the UI.

Created for portfolio and resume demonstration.