EdgeSum: Edge-based video summarization with dash cams

With billions of Internet of Things (IoT) devices, such as sensors, security cameras, and dash cams, generating huge amounts of data and transferring it to the cloud, it creates a network bottleneck with the increase of latency and bandwidth usage. Edge computing (EC) as an emerging technology is able to lighten the burden by bringing computational processes to the network edge close to data sources. According to Cisco [1], 75% of generated data consuming network bandwidth is video data. Traditionally video data is handled in the cloud due to its requirements of large storage space and high computational capacity. Dash cams are becoming prevalent as more drivers include them in their vehicles for surveillance or future incident investigation purposes. They are one representative type of IoT device that constantly generates large amounts of data. With such small storage space, the loop mechanism is a common implementation which allows the device to 'override' older video files when it has reached maximum storage capacity. In this paper, we design EdgeSum as an edge-based video summarization framework that utilizes mobile devices in the form of edge servers to summarize/compress video data of dash cams before uploading to the cloud for further processing and archiving purposes. The results support the feasibility of the framework in real-world practical applications including vehicles in driving mode, vehicles in parked mode, and surveillance applications. Based on the results, the framework delivers satisfactory performance in reducing latency and bandwidth usage by compressing the video data through summarization technique.