shakasom / Deep-Learning-for-Satellite-Imagery

Deep learning courses and projects

Deep-Learning-for-Satellite-Imagery

This repository holds some personal projects in applying deep learning to satellite imagery/Remote Sensing. My interest is in Land Cover/Land use applications and object detecion.

Land Use / Land Cover Classification

1. Austin Zoning Satellite Images - How to classify the zoning of satellite pictures of the city of Austin?

The Austin Zoning Satellite Images dataset is hosted in Kaggle. It is small dataset(3667 satellite images) with a well defined and fine grained classification labels. Training the dataset with a simple convolution Neural Network (CNN) results in almost 0.90 accuracy.

Notebook and Data availabe as Kaggle Kernel here: https://www.kaggle.com/shakasom/austin-zoning-classification-cnn-keras .

Zoning Codes Guide http://www.austintexas.gov/sites/default/files/files/Planning/zoning_guide.pdf .

Zoning district description http://www.austintexas.gov/page/zoning-districts

2. Eurosat Classification -

Object Detection