healdz/Predicting_Energy_Consumption_With_Convolutional_Neural_Networks

This repository contains code that implements a pipeline (employing raster.io, gdal, and SentinelHub API) in order to create a Convolutional Neural Network with Transfer Learning (created using TensorFlow) designed to predict the localized energy consumption based off satellite imagery.

Data_Science_Capstone

Predicting energy consumption in least developed countries with a Convolutional Neural Network and Transfer Learning.

The lights at night data set can be found at: https://eogdata.mines.edu/download_dnb_composites.html. A unique client and secret ID needs to be created via https://www.sentinel-hub.com/ in order to access the Sentinel Satelite Imagery via the pipeline created for this project.