This repository contains the work completed by me and Liam Adam for the Deep Learning course as part of the MSc Geo-Information Science at Wageningen University & Research. The project involves building a neural network from scratch using PyTorch and Jupyter Notebook to classify images from the UCM dataset. In particular, four different neural networks were used for this project: LeNet (trained from scratch), AlexNet (both from scratch and pre-trained with its default weights) and TrimoNet (a CNN written from scratch).
The repository includes a Jupyter notebook detailing the code and logic behind our neural network, as well as a PDF report discussing our findings and outcomes.
- TrimoNet.ipynb: this Jupyter notebook contains the Python code for creating, training, and testing our neural network. It provides a step-by-step walkthrough of our process, including data preprocessing, model architecture setup, model training, and model evaluation.
- report.pdf: this PDF document is a comprehensive report discussing the project objectives, methodology, and results. It offers insights into our thought process and decisions made throughout the project.