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.
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)
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 MCRInstaller_v71.zip 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
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.