Video: 360 Action via Pico Clusters

In this video, we show the challenges and directions how to find objects and actions in multi-view video 360 video streams to be processed on pico-clusters which are becoming strong candidates for edge computing. The challenge of finding objects and actions in video 360 is the equi-rectangular video representation of video 360 which causes distortions of objects and actions. Another challenge is the enormous computational needs to do the object and action detection on light-weight edge devices such as pico-clusters which are very appropriate for military settings. Our solution shows capabilities of action detection in video 360 on pico-clusters with strong accuracy which can assist in needs of detection, identification, localization and tracking of adversarial targets.