Dense VS Convolutional Neural Networks

This repository contains a Jupyter notebook which measures the effect of different types of Neural Networks on image classification.

More specifically, it performs a comparison between using a Fully-connected neural network and a Convolutional neural network for multi-class image classification.

Two sets of experiments are conducted:

• One with the use of the MNIST database.
• One with the use of the SVHN dataset.

In every experiment, for every neural network type, the following metrics are measured:

• Training time
• Accuracy
• Log-loss
• Precision
• Recall
• F1 score