Personnal project for the course Machine Learning for Earth and Environmental Sciences (UNIL, FGSE, Fall 2022)
In this assignment, I show how a supervised classification algorithm (Random Forest) applied to multispectral satellite images is used to quantify the area affected by a recent tailings dam’s collapse in South Africa.
The data consists of 2 Sentinel-2 images, one from July 13 and one from September 16 2022. They were aquired on Google Earth Engine. See "gee-script.js.
The training data was created in Qgis and is located in the shp folder.
randomforest-classification.py is the main script.