Automatic detection of native invasive rush species with aerial imagery and deep learnin

The expansion of native invasive rush species in British marginal upland grasslands results in decreased biodiversity and agricultural productivity. To successfully manage such species requires knowledge of their spatial distribution, which is costly to obtain using manual approaches but can be achieved efficiently using deep learning. Previous work documenting rush expansion adopted a manual approach. Here, we trained a U-Net deep convolutional neural network to automatically detect rush presence on aerial imagery in North-West England and demonstrate how it may be used to quantify change in rush cover over space and time.