Moderate Resolution Remote Sensing and Machine Learning for Human Rights Monitoring: The Case of Rakhine State, Myanmar

The use of machine learning on freely available remote sensing imagery has a number of applications, one of which is the monitoring of violent conflict targeting civilians. Whilst there has been significant work in using remote sensing in this way, there currently is a lack of research applying supervised learning methods to human rights specific phenomena. This paper describes a method which uses supervised learning and a convolutional neural network to detect changes in reflectance associated with the destruction of settlements and dwellings in Myanmar.