Character Recognition Using CNN for OCR

This project implements a Convolutional Neural Network (CNN) to recognize characters for Optical Character Recognition (OCR). The model is designed to classify handwritten or printed characters, providing a foundation for OCR systems.

CNN Structure used :

3 Convolution layer with 1 Dense Layer and Output dense layer to classify among the 35 Different Characters.

Prerequisites

  Python 3.8+
 TensorFlow/Keras
 Libraries: KaggleHub, NumPy, Matplotlib, Pandas, and OpenCV

Datasets used:

 For Training - https://www.kaggle.com/datasets/vaibhao/handwritten-characters
   Consists of individual chracters to train the CNN Model
  For Testing - https://www.kaggle.com/datasets/landlord/handwriting-recognition
   Words to test the Trained CNN Model

Results :

 The model was run for 70 epochs provided the following results
 Training Accuracy: 94.3%
 Validation Accuracy: 91.4%