This project implements a neural network to recognize handwritten characters, including both simple and capital letters. Given the high variance in individual handwriting styles, the system utilizes a robust training approach with the EMNIST (Extended MNIST) dataset to achieve high generalization accuracy.

Technical Approach

• EMNIST Dataset: Utilized for its comprehensive collection of handwritten digits and letters, providing a solid foundation for training.
• Preprocessing: Used OpenCV for image processing tasks such as thresholding, resizing, and normalization to match the network’s input requirements.
• Model Architecture: Built with TensorFlow/Keras, employing Convolutional Neural Layers (CNNs) to extract spatial features from character images.