Satellite Image Segmentation

This project focuses on satellite image segmentation using PyTorch, applied to a dataset of satellite images from Dubai. The goal is to segment and classify different land cover types within the images

Dataset

https://www.kaggle.com/datasets/humansintheloop/semantic-segmentation-of-aerial-imagery

Save it in data directory and name the folder dubai_datest

Code

In code directory there are the moduls for the UNet implementation

• Dataset
• Utils
• Model
• Train

Run train.py by python train.py

Notebook

The training metrics are saved in a csv and are visualized in a jupyter notebook in notebooks directory

Docs

Full project report in pdf

Group

Project made by

• Konstantinos Kafteranis (2162765)
• Jon Dorronsoro (2168329)