Abstract
There is a ccritical need to accurately represent environments volumetrically from a single or sparse set of photos for many defense related purposes. Monocular Depth Estimation (MDE) is an emerging field that seek to estimate distances from a single image. While state-of-the-art MDE algorithms are greatly improving at estimating relative depths, current models struggle to represent true depths, or metric-depths. As a result, depth estimates from these algorithms are difficult to accurately volumerically represent these results. This work seeks to explore scaling monocular depth estimation based on existing 3D information available a priori datasets. Results show that leveraging existing 3D information content can greatly improve the accuracy of MDE algorithms in a way that depth information can be used to improve volumeric characterization of new features within an area.