Goal-oriented top-down probabilistic visual attention model for recognition of manipulated objects in egocentric videos | Publicación