Title: Clothed human detection Speaker: Nathan Sprague Abstract: We present a trainable system for locating clothed people in digital images. People detection is a particularly challenging image understanding problem; as a result of variations in clothing and posture the appearance of people may vary enormously from image to image. Our approach attempts to construct a maximally person-like assembly of image regions, where candidate regions are provided by color-based segmentation followed by non-purposive grouping. A tree structured probability model is employed to allow efficient search. This structure represents the pairwise configuration of body parts as a function of relative position, relative size, and adjacency. We also consider the results of face and skin detection to narrow the search. The problem of occlusion is addressed through a mixture of trees, where the different mixture components represent the possible configurations of visible parts. In order to allow for possible over-segmentation the set of candidate image regions is augmented with pairs and triplets of adjacent regions. Experimental results will be shown to demonstrate the promise of and challenges for the current system.