deepTerra -- AI Land Classification Made Easy

TL;DR

This paper presents deepTerra, an integrated platform for end-to-end land classification using satellite imagery. It combines data collection, augmentation, training with multiple CNN architectures, and prediction with rich visualization and export capabilities, enabling researchers to build robust models with minimal effort. Through case studies on garbage detection, private pool identification, beehive localization, and cats vs dogs, the approach demonstrates the benefits of augmentation and architectural flexibility across diverse domains, achieving notable accuracies and GIS-friendly outputs. The work highlights practical impact for urban planning, environmental monitoring, and resource management, and outlines future enhancements such as multispectral analysis, segmentation, and real-time UAV support to broaden applicability.

Abstract

deepTerra is a comprehensive platform designed to facilitate the classification of land surface features using machine learning and satellite imagery. The platform includes modules for data collection, image augmentation, training, testing, and prediction, streamlining the entire workflow for image classification tasks. This paper presents a detailed overview of the capabilities of deepTerra, shows how it has been applied to various research areas, and discusses the future directions it might take.

Figures (13)

• Figure 1: deepTerra - home screen
• Figure 2: Collecting data from a large image
• Figure 3: Classifying data from image patches
• Figure 4: Image Augmentation
• Figure 5: Training a model