Crowd and people counting aims to estimate the number of people in crowded images or videos from surveillance cameras. Accurately estimating the number or density of crowds is an increasingly important application for purposes of crowd control and public safety. In overcrowding scenarios, people counting offers an essential piece of information which can be used as an accident avoidance and congestion control mechanism.
Some examples of Abnormal Crowd Behavior Detection
Example of Crowd Counting
The research behind
The network architecture basis for Abnormal Crowd Behavior
The network consists of two streams. The first one takes the density heat-maps and the second one the optical flow compared to the previous frame. Long short Term Memory networks have been used in order to learn the regular patterns from training videos. The output of each model ended up in two fully connected layers and then merged.
The network architecture basis for Crowd Counting
Inception like module which is trained with crowd images and outputs density heatmaps and crowd density values.
- L. Lazaridis, A. Dimou and P. Daras, "Abnormal Behavior Detection in Crowded Scenes Using Density Heatmaps and Optical Flow"– Submitted in 26th European Signal Processing Conference, EUSIPCO 2018, Rome September 3-7, 2018.