Distributed activity recognition in camera networks via low-rank matrix recovery

Multi-view action recognition addresses many challenges like view-invariance and occlusion. But the huge amount of data makes it hard for use in real life applications. In this paper, we propose a distributed activity classification framework based on consensus matrix completion, where several smart cameras are observing the scene, each process their observations and come to an agreement about the activity class, through communication. We validate our approach using IXMAS dataset.