Abstract

Occlusion poses a significant difficulty for object recog-nition due to the combinatorial diversity of possible oc-clusion patterns. We take a strongly supervised, non-parametric approach to modeling occlusion by learning de-formable models with many local part mixture templatesusing large quantities of synthetically generated trainingdata. This allows the model to learn the appearance ofdifferent occlusion patterns including figure-ground cuessuch as the shapes of occluding contours as well as the co-occurrence statistics of occlusion between neighboring parts. The underlying part mixture-structure also allows the model to capture coherence of object support masks between neighboring parts and make compelling predictions of figure-ground-occluder segmentations. We test the result-ing model on human pose estimation under heavy occlusion and find it produces improved localization accuracy.