Socially Constrained Structural Learning for Groups Detection in Crowd
This addresses the challenge of understanding collective behavior in crowds for applications like surveillance or social analysis, but it is incremental as it builds on existing methods with specific enhancements.
The paper tackles the problem of detecting social groups in crowds by proposing a novel algorithm that uses Correlation Clustering on trajectories, achieving state-of-the-art results with both ground truth and extracted tracklets.
Modern crowd theories agree that collective behavior is the result of the underlying interactions among small groups of individuals. In this work, we propose a novel algorithm for detecting social groups in crowds by means of a Correlation Clustering procedure on people trajectories. The affinity between crowd members is learned through an online formulation of the Structural SVM framework and a set of specifically designed features characterizing both their physical and social identity, inspired by Proxemic theory, Granger causality, DTW and Heat-maps. To adhere to sociological observations, we introduce a loss function (G-MITRE) able to deal with the complexity of evaluating group detection performances. We show our algorithm achieves state-of-the-art results when relying on both ground truth trajectories and tracklets previously extracted by available detector/tracker systems.