Abstract: The accurate delineation of crop fields is essential for better decision-making by farmers and workers in the agricultural sector by facilitating land registration and acquisition. Crop field boundary data can also help inform climate change policies by providing better estimates of cropland area for GHG estimation and damage assessment after a natural disaster.
In this tutorial, you will learn how to implement crop field delineation by training a computer vision model for semantic segmentation using satellite images. Specifically, you will be creating a Pytorch implementation of the U-net architecture, downloading your own Sentinel-2A satellite images using Google Earth Engine, and using the trained model to automatically generate crop field maps for a region of interest. This tutorial is aimed at data scientists with prior background in deep learning looking for concrete examples on the application of deep learning to tackle a problem related to climate change and agriculture.
Bio: Bio Coming Soon!
Machine Learning Researcher | Thinking Machines Data Science