Transferable Crowd Counting and Localization for Urban Operations

In urban operations, an intuitive common operational picture (COP) is essential for gathering necessary information for operators like fire fighters, police, and military personnel. This work emphasizes accurate crowd counting and localization in urban settings using video surveillance and AI, particularly deep learning with vision transformers. Results are visualized as a 2D human density layer in the COP. Addressing data set variability, the presented multiresolution approach enables transferable AI models, overcoming the challenge of diverse training and realworld data.