ESCal: Efficient and Scalable Calibration of Camera Networks Using a Top View

Calibrating camera networks is a complex yet crucial task for ensuring accurate data collection, particularly for large-scale scenarios. This work's main contribution is a method for efficiently estimating the network's extrinsic paramaters for every camera. It also determines their effective coverage without requiring overlapping fields of view. We propose ESCal, an efficient and scalable solution that leverages homographies and point correspondences while disregarding their prerequisite of planar scenes. It semi-automatically determines the pose of the considered cameras given one top view of the scene. Our approach is designed to be fast, easy to use, scalable, and sufficiently accurate. We demonstrate these qualities under real-world conditions.