Semantically accurate segmentation of a particular Object Of Interest (OOI) in an image is an important but challenging step in computer vision tasks. Our recently proposed object-specific segmentation algorithm learns a model of the OOI which includes information on both the visual appearance of and the spatial relationships among the OOI components. However, its performance heavily depends on the assumption that the visual appearance variability among OOI instances is low. We present an extension to our algorithm that relaxes this assumption by incorporating shape information into the OOI model. Experimental results and an ANOVA-based statistical test confirm that the incorporation of shape has a highly significant positive effect on segmentation performance.