N-best maximal decoders for part models

12 years 4 months ago

Download www.ics.uci.edu

We describe a method for generating N-best conﬁgurations from part-based models, ensuring that they do not overlap according to some user-provided deﬁnition of overlap. We extend previous N-best algorithms from the speech community to incorporate non-maximal suppression cues, such that pixel-shifted copies of a single conﬁguration are not returned. We use approximate algorithms that perform nearly identical to their exact counterparts, but are orders of magnitude faster. Our approach outperforms standard methods for generating multiple object conﬁgurations in an image. We use our method to generate multiple pose hypotheses for the problem of human pose estimation from video sequences. We present quantitative results that demonstrate that our framework signiﬁcantly improves the accuracy of a state-of-the-art pose estimation algorithm. We address the task of generating multiple candidate object conﬁgurations in an image or video, within the framework of part-based models. Su...

Dennis Park, Deva Ramanan

Real-time Traffic