Output details
15 - General Engineering
University of Oxford
Progressive search space reduction for human pose estimation
The objective here is to estimate 2D human pose as a spatial configuration of body parts in an image. Apart from the novelty of the approach (which combines detection, segmentation and tree graphical models), the paper was one of the first to successfully estimate human pose in very visually challenging (for a computer) images from TV and film. It also introduced a benchmark data set and evaluation procedure for assessing the performance of human pose estimation algorithms, see http://www.robots.ox.ac.uk/~vgg/data/stickmen/. This data set and evaluation protocol have been used extensively by other researchers. A journal version was published in 2012.