Performance and Privacy Aspects of Image Classification in Cross-platform Mobile Applications

Mobile devices and applications that run on them have changed the everyday lives of not just technology professionals but of the public at large. Another technology that is introducing a rapid revolution in our lives is the use of machine learning in advanced applications, enabling use cases that were deemed unfeasible a short time ago. The fact that mobile devices are resource-constrained plays a crucial role, as does the fact that the platform landscape is not homogenous and cross-platform application development is very popular. This paper presents an analysis of the problem of privacy and performance in the use of a machine learning approach for image classification in cross-platform mobile applications. Two major approaches and three different metrics are analyzed within the scope of their use for example privacy honoring mobile application that allows tourists to identify landmarks from a photograph.