This project implements a Convolutional Neural Network (CNN) to classify handwritten digits from the MNIST dataset with high accuracy.
CNNs are widely used for image recognition tasks due to their ability to capture spatial hierarchies in data.
This repository provides the full pipeline for training, evaluating, and visualizing CNN performance on MNIST.
The MNIST dataset consists of 60,000 grayscale images (28x28 pixels) of handwritten digits (0-9).
Despite its simplicity, MNIST is a benchmark dataset in deep learning, often used for testing new neural network architectures.
CNNs are particularly effective for MNIST due to their ability to learn spatial hierarchies in image data.
This project aims to:
✔ Train a CNN from scratch on MNIST
✔ Optimize the model for high accuracy and fast convergence
✔ Compare different architectures and visualize the results