Gatete8/RTDS-Radical-terrace-detection-system-

RTDS — Radical Terrace Detection System: A GeoAI workflow that converts aerial orthophotos into terrace maps, statistics, and plain-language erosion insights. Built with R, terra, sf, Leaflet, and Random Forest. Open source · Rwanda-focused · globally relevant—connecting pixels to soil conservation decisions. # RTDS — Radical Terrace Detection System **Interactive web application for automated terrace detection from orthophotos, with map visualization and statistical reporting.** --- ## Live demo Try the app in your browser (no installation required): **bugingo.shinyapps.io --- ## Introduction **RTDS** (Radical Terrace Detection System) is a Shiny-based workspace for mapping **agricultural terraces** from high-resolution **GeoTIFF orthophotos**. It targets users who need repeatable, map-based insight for planning, research, and field programs—planners, agronomists, and GIS analysts working with aerial or drone imagery. The app guides you through upload, runs a **Random Forest** classifier on engineered spectral and textural features, and presents **terrace polygons** on an interactive **Leaflet** map alongside **coverage**, **area**, **feature count**, and **confidence** summaries, plus a plain-language **interpretation** of erosion-related risk. --- ## Problem this project addresses Terraced landscapes are a proven soil and water conservation measure. Mapping where terraces exist—and how much of a study area they cover—is often done manually or with ad hoc tools, which is **slow**, **hard to reproduce**, and **difficult to communicate** to decision-makers. RTDS addresses this by: - **Automating** terrace delineation from a user-provided orthophoto (within the limits of the trained model and input quality). - **Standardizing** outputs: georeferenced results, consistent statistics, and a shareable web workflow. - **Making results accessible** through a map-first interface and readable metrics, so technical and non-technical stakeholders can use the same view. The system is designed for **Rwanda-relevant** workflows (as reflected in the app title and documentation) but can be used wherever similar orthophoto data and model assumptions apply. --- ## Features - **Landing page** with guided introduction and entry to the main workspace. - …