Historical Stereogram Depth Finding is a computer vision and machine learning application that works to restore the names of Civil War soldiers buried in unmarked graves at St. Augustine National Cemetery. Sponsored by Dr. Amy Giroux and the National Cemetery Administration, the project confronts a problem that can’t be solved by conventional means: federal regulations prohibit disturbing the gravesites with radar or excavation, so the only path to identifying these graves runs through the historical record itself—specifically, 19th-century stereographic photographs that captured the original wooden grave markers before they were lost. The software analyzes these stereo image pairs to reconstruct the 3D layout of the scene, measuring real-world distances and angles between the graves shown in the photos. Because the specifications of the original cameras are unknown, the team developed a machine learning model to estimate critical parameters like focal length and field of view, then built a depth-finding pipeline to translate the flat historical images into spatial measurements. Researchers can plot points of interest directly on a photograph and have the system calculate the corresponding real-world positions—pinpointing which of today’s “unknown” headstones belongs to which named soldier. The result is a non-invasive way to give long-anonymous graves their names back, honoring the fallen while preserving the sites that hold them.
