Composite Augmentations for Semantic Segmentation in Aerial Images with Few Samples

Remote sensing and computer vision have the potential to enable comprehensive population monitoring to inform wildlife and biodiversity conservation. However, annotated datasets of wildlife in-situ are often difficult, expensive, and time-consuming to procure. This paper proposes a computational and data efficient method to synthesize composite images to supplement real-world data in data-sparse environments with few positive samples. We evaluated our method on three aerial remote sensing datasets and demonstrated a 3% increase in target-class IoU scores. We aim to use this method with a novel aerial dataset of the Boreal forest for ungulate monitoring, which is presently under development.