Using a Real-Time Object Detection Application to Illustrate Effectiveness of Offloading and Prefetching in Cloudlet Architecture

XuTong Zhu, The University of Western Ontario

Abstract

In this thesis, we designed and implemented two versions of a real-time object de- tection application: A stand alone version and a cloud version. Through applying the application to a cloudlet environment, we are able to perform experiments and uses the results to illustrate the potential improvement that a cloudlet architecture can bring to mobile applications that require access to large amounts of cloud data or intensive com- putation. Potential improvements include data access speed, reduced CPU and memory usages as well as reduced battery consumption on mobile devices.