This software produces confidence images of geometric classes and attempts to create a 3D pop-up model based on the geometry. This software is made available for dissemination and research comparison purposes without any warranty. Commercial use is prohibited. See the README and/or LICENSE.txt file for more information. The executable takes in an original image, a superpixel image, and learned models of geometry (provided). It outputs a labeled image, confidence maps for labels, and the vrml files. If you obtain nice results that you would like to share, we would be happy to see them.  Source code is now provided.

Please note that this software is primarily intended for researchers, and I can provide only minimal support.

The MCRInstallers are copyright Mathworks and distributed exclusively for use with the photo pop-up program as provided by the Matlab license agreement section 5.1.1.

View this page in Romanian courtesy of azoft.
View this page in Danish courtesy of Einar Solbakken.

Source Code (New!)

This source code includes the ICCV 2005 (Geometric Context), IJCV 2007 (Surface Layout), and SIGGRAPH 2005 (Automatic Photo Pop-up) code.

Matlab source code  (3.8MB)


Here is more general code for learning (and applying) region-based classifiers given images with pixel labels for training.

Matlab source code


Original Version

This is a post-ICCV (10/07/05) version of the code and may produce slightly different (hopefully a little better) results than were described in "Geometric Context from a Single Image". ICCV and SIGGRAPH executables may be available to academic researchers upon request for the sake of comparison.

MATLAB 7 SP3 libraries (100 MB): Download if you do not have MATLAB 7 SP3 installed or have trouble with libraries.
Detailed Instructions courtesy of Marion Bates

MATLAB 7.04 libraries (100 MB): Download if you do not have MATLAB 7.04 installed or have trouble with libraries.
Some have trouble compiling the segment superpixel code from Felzenszwalb, so I have put it here: segment.exe (from Rafael Cappa)

IJCV Version

This version uses the latest outdoor and indoor classifiers, described in our IJCV paper "Recovering Surface Layout from an Image". This software is intended to be used as a research tool and for comparison in academic research. Commericial use is prohibited. Included is a file that must be unzipped, and path variables must be set appropriately.
Linux: Photo Pop-up IJCV version

Windows version below.  See README for installation procedures.  Note that this version is very slightly different than the Linux version, offering similar performance but much faster.

Windows: Photo Pop-up IJCV version


Some of this material is based upon work supported by the National Science Foundation under CAREER Grant No. ISS-0546547. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. This work is also partially supported by a Microsoft Research Fellowship awarded in 2006.