In this paper a novel framework for three-dimensional surface reconstruction by self-consistent fusion of shading and shadow features is presented. Based on the analysis of at lea...
Most state-of-the-art approaches to action recognition rely on global representations either by concatenating local information in a long descriptor vector or by computing a single...
While great strides have been made in detecting and localizing specific objects in natural images, the bottom-up segmentation of unknown, generic objects remains a difficult chall...