Part Affinity Field based Activity Recognition

This report presents work and results on Activity Recognition using Part AffinityFields for real-time surveillance applications. Starting with a short introductionto the motivation, this report gives a detailed overview over the key idea ofthe pursued approach and explains the basic ideas. In addition a variety ofexperiments on various subjects are presented, like i) the impact of the numberof input frames, ii) the impact of different simple dimensionality reductionapproaches, and iii) a comparison on how multi-class and binary problemformulation influence the performance.