MarshCover: Vegetation Density Estimator for Marsh

This app is used to estimate either the densities of vegetated and unvegetated points in marsh graound cover images, or the densities by species within the image. The app identifies densities in the image ofSpartina alterfloris, Juncus gerardii, Batis maritima, Sarcocornia perennis, Avicennia germinans, and unvegetated points.

Given an image, a specified number of points, and a selection of either binary or full estimation, the system randomly selects those points in the image to create 33x33 snippets around those points and pass them to the trained AI models for classification to estimate densities. A count of each class is kept and displayed on the result page as a pie chart with either two or six classes, depending on the estimation choice being binary or full, respectively. Snippets, coordinates of chosen points and their predicted labels are presented too.

Instructions

To use this app, pick a marsh groundcover image for analysis. This image must be at least 33x33 and must be a full color (rgb) image. After an image is chosen, specify the number of points you would like analyzed along with whether you want Binary Estimator (two classes: unvegetated and vegetated) or Full Estimator (six classes: unvegetated and the five species mentioned above). The number of points chosen must be a positive integer. Finally click the submit button to get the analysis.