MNIST Digit Classification with Keras and K-Fold Cross Validation

Project Overview

This project focuses on building and training a Multi-Layer Perceptron (MLP) to classify digits from the MNIST dataset (0-9). The implementation uses TensorFlow and Keras libraries. The project includes:

• Data Augmentation: to generate more diverse training samples using transformations such as rotation, shifting, and zooming.
• K-Fold Cross Validation: to evaluate the model performance across different subsets of the data and ensure robustness.
• Model Visualization: to track training accuracy, validation accuracy, and loss metrics over epochs.

Authors

• Szymon Szulc - ID: 21323208
• Boris Bobylkov - ID: 21317097

---

Prerequisites

• Python 3.x
• TensorFlow 2.x
• Matplotlib
• Scikit-learn