F-Formation Detection: Individuating Free-Standing Conversational Groups in Images

摘要: Detection of groups interacting people is a very interesting and useful task in many modern technologies, with application fields spanning from video-surveillance to social robotics. In this paper we first furnish rigorous definition group considering the background sciences: allows us specify kinds group, so far neglected Computer Vision literature. On top taxonomy present detailed state art on detection algorithms. Then, as main contribution, brand new method for automatic still images, which based graph-cuts framework clustering individuals; particular, are able codify computational sense sociological F-formation, that encode having only proxemic information: position orientation people. We call proposed Graph-Cuts F-formation (GCFF). show how GCFF definitely outperforms all methods terms different accuracy measures (some them new), demonstrating also strong robustness noise versatility recognizing various cardinality.