Demo Abstract: On-Demand Information Retrieval from Videos Using Deep Learning in Wireless Networks

Mobile devices with cameras have greatly assisted in the prevalence of online videos. Valuable information may be retrieved from videos for various purposes. While deep learning enables automatic information retrieval from videos, it is a demanding task for mobile devices despite recent advances in their computational capability. Given a network consisting of mobile devices and a video-cloud, mobile devices may be able to upload videos to the video-cloud, a platform more computationally capable to process videos. However, due to network constraints, once a query initiates a video processing task of a specific interest, most videos will not likely have been uploaded to the video-cloud, especially when the query is about a recent event. We designed and implemented a distributed system for video processing using deep learning across a wireless network, where network devices answer queries by retrieving information from videos stored across the network and reduce query response time by computation offload from mobile devices to the video-cloud.