A Content-based Intelligent Mobile Video Censoring System

With the development of information technology, mobile video on-demand and live streaming prevails in the young population. But the content of mobile video may contain some bloody and violent scenes. Teenagers, among all consumers, are more susceptible the negative influence of violence in videos. The censorship targeting at the online contents including images and video is an important yet arduous task. The adoption of AI-based censoring technique can help lower the operation cost for Internet content providers and save the work labor in censorship while preventing video consumers from being affected by unhealthy information. This paper presents an embedded system for content-based video censoring based on the deep learning technique. The proposed system is capable of filtering video content automatically, where an algorithm is developed that utilizes the power of Intel IoT development suite and the Intel Neural compute stick. Efforts has been made to optimize the real-time content censoring of the system. Experimental results show that, when applied to the video-on-demand IoT applications, the proposed system can speed up the video processing significantly and it can be effective in shielding young audience from violent videos at mobile devices in a real-time manner.