Adaptation of Machine Learning Frameworks for Use in a Management Environment

The combination of person and location recognition provides numerous new fields of possible applications, such as the development of approaches for detecting missing persons in public spaces using real-time monitoring. It is necessary to use frameworks of both domains on given data sets and to merge the acquired results. For this purpose Thomanek et al. [11] developed an evaluation and management system for machine learning, which allows the interconnection of different frameworks and the fusion of result vectors [11]. This paper discusses the EMSML in terms of interfaces and components to develop a generic workflow that supports the integration of different frameworks for people and location recognition. In this context, the focus is on the required adaptation of existing frameworks to the implemented infrastructure. A generic workflow concept can be deduced from the analysis results. This concept can be applied to two typical frameworks for evaluation and implemented as prototypes. Subsequently, developed test cases are used to demonstrate the functional validity of the prototypes and the applicability of the concept.