MLP-and-CNN-for-Image-Classification

The objective of this project was to develop a Multilayer Perceptron (MLP) model for multiclass classification of images from the widely used CIFAR-10 dataset. Additionally, a pre-trained Convolutional Neural Network (CNN) was employed, and the performances of both models were compared. The MLP model was constructed from scratch to allow for step by step analysis of a feedforward neural network’s architecture. This allowed to explore different implementations for backpropagation, analyze how the depth of a neural network influences its performance, and how various network architecture choices affect the model’s quality, as we will discuss in this paper. This report discusses the effect of implementing regularization techniques, such as L2-regularization, on the accuracy of the model. Furthermore, we investigated how ML algorithms can be used to address computer vision problems and image data treatment. We examined the differences between how MLP and CNN treat data and the resulting implications.