Part Affinity Field based Activity Recognition

This report presents work and results on Activity Recognition using Part Affinity Fields for real-time surveillance applications. Starting with a short introduction to the motivation, this report gives a detailed overview over the key idea of the pursued approach and explains the basic ideas. In addition a variety of experiments on various subjects are presented, like i) the impact of the number of input frames, ii) the impact of different simple dimensionality reduction approaches, and iii) a comparison on how multi-class and binary problem formulation influence the performance.