Towards edge-caching for image recognition

摘要: With the available sensors on mobile devices and their improved CPU storage capability, users expect to recognize surrounding environment provide relevant information and/or content automatically immediately. For such classes of real-time applications, user perception performance is key. To enable a truly seamless experience for user, responses requests need be provided with minimal user-perceived latency. Current state-of-the-art systems these applications require offloading data cloud. This paper proposes an approach allow users' onboard leverage resources closer home, i.e., at edge network. We propose use edge-servers as specialized caches image-recognition applications. develop detailed formula expected latency cache that incorporates effects recognition algorithms' computation time accuracy. show that, counter-intuitively, large sizes can lead higher latencies. best our knowledge, this first work models compute-intensive