This document summarizes a hierarchical object-based image analysis approach used to classify sub-meter multispectral imagery in Tanzania. It describes the workflow, which includes image segmentation, feature selection, classification using decision trees and random forests, and accuracy assessment. Results showed overall accuracy above 85% for both algorithms, demonstrating the effectiveness of the object-based approach for discriminating land cover classes at a detailed scale. The study was funded by Conservation International and performed using tools like eCognition and WorldView2 satellite imagery.