Minecraft, the best-selling video game of all time, is a 3D voxel-based video game where players can build, destroy, and explore an infinite, procedurally generated world. The terrain is generated using Perlin noise to produce (mostly) smooth terrain and then structures and other interesting features are placed in the world post-generation. This aim of this project will be to learn and apply tools from machine learning and deep learning such as semantic segmentation in order to identify these features and annotate the given overhead 2D map.