Human context: modeling human-human interactions for monocular 3d pose estimation

摘要: Automatic recovery of 3d pose multiple interacting subjects from unconstrained monocular image sequence is a challenging and largely unaddressed problem. We observe, however, that by tacking the interactions explicitly into account, treating individual as mutual "context" for one another, performance on this problem can be improved. Building observation, in paper we develop an approach first jointly estimates 2d poses people using multi-person extension pictorial structures model then lifts them to 3d. illustrate effectiveness our method new dataset dancing couples videos dance competitions.