Viewpoint Selection for Human Actions

Abstract

In many scenarios a dynamic scene is filmed by multiple video cameras located at different viewing positions. Visualizing such multi-view data on a single display raises an immediate question-which cameras capture better views of the scene? Typically, (e.g. in TV broadcasts) a human producer manually selects the best view. In this paper we wish to automate this process by evaluating the quality of a view, captured by every single camera. We regard human actions as three-dimensional shapes induced by their silhouettes in the space-time volume. The quality of a view is then evaluated based on features of the space-time shape, which correspond with limb visibility. Resting on these features, two view quality approaches are proposed. One is generic while the other can be trained to fit any preferred action recognition method. Our experiments show that the proposed view selection provide intuitive results which match common conventions. We further show that it improves action recognition results.

Multi-view Recognition Using Weighted View Selection

“…While complex 3D models often achieve high recognition rates, they also tend to be computationally expensive. A recent hybrid 2D/3D approach determines the best viewpoint of an action in isolation after constructing a 3D model [7]. This approach differs from ours in that it is focused on finding the single best view from the perspective of a human observer, while our method learns how to combine views to improve recognition.…”

A Data Fusion Perspective on Human Motion Analysis Including Multiple Camera Applications

“…Different utility measures have been proposed studying the saliency, concavity or variations of silhouette stacks [20]. The main drawback of this approaches is that they do not exploit the complementary information that might be present at each view.…”

An adaptively emerging mechanism for context-aware service selections regulated by feedback distributions

“…Work given in [15] is about prototype selection and [16] is concerning about view selection. Prototype selection is an important task for classifiers since through this selection process the time for classification or training could be reduced.…”

“…It reported that the experimental results showed the performance of the method and compared accuracy and runtimes against other prototype selection methods. They automated a process by evaluating the quality of a view, captured by every single camera, for which a human producer manually selects the best view in [16]. They regard human actions as threedimensional shapes induced by their silhouettes in the space-time volume.…”