• Mobile Cloud Computing
• Parallel Systems
• Computer Vision
• Object Instance Recognition
• Cloud Computing

Cloud Computing made the idea of on-demand, scalable and high available resources a low-cost reality. In the smartphone era, where devices (with limited computation and battery power) produce a massive quantity of data, Mobile Cloud Computing has become an hot topic in research community. In particular, it has been successfully exploited in the development of many Object Instance Recognition (OIR) applications designed for mobile devices, where local computation is too much expensive for the single smartphone. In this context, reduce the time needed for tasks computation gain advantages for both providers and end-users, giving a better user experience and saving battery life. In this paper we introduce CloudCache, a framework designed to improve performance of mobile OIR-related applications, where previous (expensive) computations performed by the Back End system are reused in order to speedup similar tasks in the future. The design and implementation of CloudCache is described, where different tools from previous works in Computer Vision has been used to achieve the desired performance. Some of these tools has been parallelized to exploit the multi-cores architectures commonly presented in most of Cloud providers nowadays, obtaining real-time results. CloudCache has been tested on three recognition applications, where two new image datasets are presented to evaluate the system, providing results for similar tasks previously computed in 13 milliseconds in the worst case on a six cores machine.