Deep-Learning

Assignment on Deep-Learning course at ECE AUTH

Deep Learning Assignment – ECE AUTH

This repository contains my assignment for the Deep Learning course at the Department of Electrical and Computer Engineering (ECE), Aristotle University of Thessaloniki (AUTH).

The goal of the assignment is to study and compare different deep learning architectures and training strategies on a small medical imaging dataset.

Contents

The assignment consists of three main parts:

1. CNN

   • Implementation of a Convolutional Neural Network and experimenting with different techniques

2. Transfer Learning with CNN

   • Transfer learning using a pretrained ResNet18 model
   • Comparison between feature extraction and fine-tuning

3. Transfer Learning with Vision Transformer

   • Transfer learning using a pretrained DeiT Transformer
   • Compoarison between feature extraction and fine-tuning

Dataset

The dataset used in this project is BloodMNIST, from the MedMNIST collection.

Framework

• PyTorch
• timm (for pretrained Vision Transformer models)

---