Study on interactive face recognition

In recent years, there have been significant advances in the field of face recognition. Many of them are
based on deep learning. In our thesis, we investigated the state of the art face recognition technology
and its possible applications. The main goal of our work is to build a prototype for the exhibition
“Informatik zum Anfassen”.
At the beginning, we evaluated different classification technologies (OpenCV, Microsoft Cognitive
Services Face API, TensorFlow) and their capabilities. On the results we gathered, we built an early
prototype for classification from phones. We changed direction and developed the second prototype
that is capable of fully autonomous tracking.
Our prototype demonstrates with a full-stack example how embeddings can be used to train an
unsupervised online clustering algorithm to track previously unknown people over different locations.
To accomplish this, we developed a client-server architecture comprising multiple services to achieve
high performance and scalability.
We are confident that we have achieved the goal of our thesis with our prototype. It shows the
capabilities of modern face recognition technology by applying them effectively to a real world setting.
In our opinion it would be a worthwhile addition to the exhibition.