Visual Crowd Analysis / Fire Protection Directorate

At major events, such as concerts in the Munich Olympic Hall, it is important for security reasons to be able to estimate or determine the number of people in a crowd or in a certain area quickly and as accurately as possible. This is necessary so that emergency services and security personnel on site can control the inflow and outflow of people or the utilization of certain areas (e.g. in so-called breakwaters) in order to avoid exceptional situations and comply with applicable safety guidelines.

Manual or human estimation of crowd density or counting people in a crowd often proves to be difficult, inaccurate and too slow to react quickly to changes.

A machine and automated solution for counting people in certain areas would greatly facilitate the work of the fire protection directorate of the City of Munich.

Students on the Deep Learning in Visual Computing master's course are working on solutions with the support of their professor and employees of the fire department.

The results will be published here at the end of the semester.