EuroSAT Classification with PyTorch

This repository contains the implementation for Applied Data Science (ADRS) Lab 4, focusing on satellite image classification using convolutional neural networks with PyTorch.

Project Overview

This project implements deep learning models to classify satellite imagery from the EuroSAT dataset. The EuroSAT dataset contains Sentinel-2 satellite images covering 13 spectral bands and consisting of 10 classes with a total of 27,000 labeled and geo-referenced images.

Dataset

The project uses the EuroSAT dataset, which includes the following land use and land cover classes:

• Annual Crop
• Forest
• Herbaceous Vegetation
• Highway
• Industrial Buildings
• Pasture
• Permanent Crop
• Residential Buildings
• River
• Sea Lake

Requirements

The project requires the following Python packages:

torch
torchvision

Install dependencies using:

pip install -r requirements.txt

Getting Started

1. Extract the dataset: Run the unzip.ipynb notebook to extract the EuroSAT dataset from the ZIP file:

   # The notebook will extract EuroSAT.zip to a 'dataset' folder

2. Run the main implementation: Open and execute adrs_lab4/pytorch_convnets.ipynb to:

   • Load and preprocess the EuroSAT dataset
   • Implement convolutional neural networks
   • Train and evaluate the models
   • Analyze classification results

Implementation Details

The main implementation in pytorch_convnets.ipynb covers:

• Data Loading: Loading and preprocessing the EuroSAT satellite imagery
• Model Architecture: Implementation of CNN architectures for image classification
• Training: Training loop with appropriate loss functions and optimizers
• Evaluation: Model evaluation and performance metrics
• Visualization: Results visualization and analysis