Browsing by Author "Banzhaf, Clint"

Filter results by typing the first few letters
Now showing 1 - 2 of 2
  • Thumbnail Image
    ItemOpen Access
    Extracting facial data using feature-based image processing and correlating it with alternative biosensors metrics
    (2017) Banzhaf, Clint
    Extracting facial data has recently drawn more attention due to an ongoing progress in research and increase in its accuracy. Thanks to mobile platforms, the necessary but ordinary cameras are remarkably widespread nowadays. Big companies, like Google, have entered the market with their own cloud-based and mobile face analysis solutions. Possible applications for extracted facial data range from more user-friendly devices to user-interface evaluation and marketing. This thesis starts with an introduction to the topic of facial data extraction. It outlines several interesting applications, important techniques and existing frameworks first. Afterwards, the main contribution is presented: OpenFace++, an improved and extended variant of the existing OpenFace face analysis framework. Like OpenFace, the proposed OpenFace++ framework aims to close an unsatisfying gap in available functionality between commercial and free open-source solutions. OpenFace++ adds several new features to OpenFace. Amongst others, the added highlights are facial expression recognition, attention estimation, and cross-platform compatibility. OpenFace++ works on Android out-of-the-box. The work is completed by presenting a user-study which was conducted to investigate the reliability of extracted facial data. The first experiment evaluates the detectability of eye-closeness states. Results show that facial data extraction can outperform competing solutions. The second experiment shows that face orientation extraction also works reliably compared to a gyrometer. Finally, the third experiment demonstrates that the new facial expression detection can detect many cases very accurately. The thesis finishes with some suggestions for further improvements on OpenFace++ in the future.
  • Thumbnail Image
    ItemOpen Access
    Smartphone-Campusnavigation
    (2011) Banzhaf, Clint; Exner, Daniel; Lehmann, Marcus
    Diese Fachstudie untersucht mögliche Szenarien, um eine am VIS entstandene Navigations-Software von einem herkömmlichen Desktop-System auf eine gängige mobile Plattform zu portieren. Dabei werden dafür relevante Merkmale der bestehenden Software erläutert und Besonderheiten der in Frage kommenden mobilen Plattformen vorgestellt. Es ist ein detailliert herausgearbeitetes Ratingsystem für die derzeit auf dem Markt konkurrierenden mobilen Plattformen enthalten und ebenso eine daraus im Kontext der Anforderungen resultierende Empfehlung für einen dieser Kandidaten. Im Anschluss wird bereits der Grundstein für ein möglicherweise folgendes Entwicklungsprojekt gelegt, indem denkbare Use-Cases auf Grundlage von erhobenen Umfragewerten modelliert und ein Prototyp mit ausgewähltem Funktionsumfang implementiert und dokumentiert werden.