Living shoreline restoration uses natural and manmade materials to stabilize shorelines, offering a cost-effective and ecologically beneficial alternative to traditional hardened structures. Accurate and frequent monitoring is essential to track project success, a task increasingly facilitated by Unoccupied Aerial Vehicles (UAVs), which provide high-resolution imagery for precise vegetation mapping at a significantly lower cost than traditional surveys. This mapping is critical as marsh species community composition is a sensitive indicator of wetland health; shifts in dominant species signal changes in essential services like carbon sequestration and shoreline stabilization. To support monitoring at Living Shoreline sites in Alabama, this study used Random Forest (RF) classification on high-resolution UAV imagery, incorporating multispectral bands, a Digital Surface Model, and textural features to classify a marsh community comprised of several marsh taxa plus unvegetated terrain. The RF model achieved a high overall accuracy on test pixels, with the DSM and the Normalized Green Blue Difference Index identified as key predictive variables. To validate the RF classification, a ground-truthing strategy used stratified random sampling within "shrunken" polygons to ensure points were > 1 m from adjacent classes. Field data collection at each point includes: species composition, dominant cover type, canopy heights, elevation, and shoot density (collected at select points). The resulting ground-truth data will be directly compared to the RF classification to evaluate classification accuracy. The final output provides spatial coverage and total area estimates for key marsh vegetation classes, enabling the tracking of community compositional changes over time and an essential assessment of the restoration project's long-term efficacy.