A good model of object shape is essential in application such as segmentation, object detection, inpainting and
graphics, etc.
In this paper, we use a type of Deep Boltzmann Machine that we call a Shape Boltzmann Machine (ShapeBM) for the task of modeling binary shape images.
This paper define a strong model should meets two requirements:Realism/Generalization.
Previous approaches to modeling 2D shape:1.Grid MRF/CRF 2.PCA/Factor Analysis 3.template shapes 4.Shape fragments.
The main contribution of this paper is to show how a strong model of binary shape can be constructed using a form of DBM,We demonstrate that a ShapeBM trained on a relatively small dataset is both able to generate realistic samples and to generalize to generate samples that differ form images in the training dataset.
/Files/chenpingjun1990/cvpr_12_Shape_Boltzmann_Machine.pdf
/Files/chenpingjun1990/cvpr_12_eslami_shapebm_code.zip"The Shape Boltzmann Machine: a Strong Model of Object Shape"
Eslami, S.M.,
Heess, N.,
Winn, J.Computer Vision and Pattern Recognition (CVPR), June 2012.