Creating a dataset of satellite images for StyleGAN training

Introduction

Training a first GAN

Defining data qualities

  1. Satellite imagery (obviously) of the RGB color channels
  2. Diverse in scenery (to hopefully have a similar diverse output after the training process)
  3. 1024x1024 pixel in resolution (based on reference resolutions commonly used with StyleGAN)
  4. Visually appealing (a big learning from the initial test training with the AID dataset)
  5. Ideally more than 2000 images (nothing to back up that number, except what seems to be the minimum required for StyleGAN training)

Searching for open data sources

In tiny font at the bottom, you can see the data sources Google Maps is using for this specific earth segment

Building a data collection pipeline

  1. SENTINEL-2
  2. QGIS
  3. Google Earth Engine
  4. Google Colab

SENTINEL-2

Google Earth Engine

QGIS

Google Colab

Tying all components together in one workflow

Cleanup the data

Generating novel images with a GAN

All those landscape images do not exist
Example of an interpolation video from one landscape to the other

--

--

--

Designer, Maker and Researcher, currently located in The Netherlands.

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Neural Networks

Testing Areas and the Training Workflow

Types of Ensemble methods in Machine learning.

Deep Dive into Computer Vision with Neural Networks — Part 1

Image classification tutorials in pytorch-transfer learning

This article was translated automatically — why I wouldn’t be on medium without DeepL

Cross Validation Explained

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Frederik Ueberschär

Frederik Ueberschär

Designer, Maker and Researcher, currently located in The Netherlands.

More from Medium

Intro to Satellite Image Processing- Part 1

Analysing an Aerial LiDAR Point Cloud dataset

AI in Agriculture: Opportunities and Challenges

AGRICULTURE CROPS IMAGES CLASSIFICATION