Code

Ceres Solver

Ceres Solver is a BSD licensed, general purpose nonlinear least squares solver with extensive support for bundle adjustment problems. It has a simple expressive API with support for automatic differentiation, local parameterizations, robust loss functions, derivative checking, multithreading and more. It comes with multiple linear solvers and implements both classical and new ideas for solving bundle adjustment problems.

For questions, bug reports, ideas for further development, please send email to the mailinglist.

Practical Global Optimization for Multiple View Geometry

Versions 1.0 [tgz]

This toolbox implements the algorithm described in

Practical Global Optimization for Multiview Geometry [pdf][code]

Fredrik Kahl, Sameer Agarwal, Manmohan Chandraker, David Kriegman and Serge Belongie
International Journal of Computer Vision, Vol. 79(3), pp. 271-284, 2008.

Examples of usage of the toolbox are in the files example_tri.m and example_res.m which de scribe how to do triangulation and resectioning respectively. You will need a recent version of SeDuMi to use this toolbox.

History

20/12/2007 Version 1.0 (initial release)

Generalized Non-metric Multidimensional Scaling

Versions 1.0 [tgz]

This toolbox implements the algorithm described in

Generalized Non-metric Multidimensional Scaling [pdf]

Sameer Agarwal, Josh Wills, Lawrence Cayton, Gert Lanckriet, David Kriegman and Serge Belongie
AISTATS 2007, San Juan, Puerto Rico.

Examples of usage of the toolbox are in the files example.m which compares the NMDS algorithm with the classical multidimensional scaling algorithm on the 10 cities dataset. You will need a recent version of SeDuMi to use this toolbox.

History

20/12/2007 Version 1.0 (initial release)

Structured Importance Sampling of Environment Maps

Versions 1.0 [tgz]

This toolbox implements the algorithm described in

Structured Importance Sampling of Environment Maps [pdf] [avi]

Sameer Agarwal, Ravi Ramamoorthi, Serge Belongie and Henrik Wann Jensen
SIGGRAPH 2003, San Diego, CA, TOG 22(3):605-612, July 2003.

You will need SCons to build it. Once you have SCons installed, invoking "scons" in the main directory will result in the binary "stratify" to be created in the ./bin directory. The program takes input in the form of .float files which are rgb float dumps in an angular uv format. Paul Debevec's site contains maps in this format. Also the program pvalue that comes with the package radiance, can create these files from .hdr formatted files.

There are a couple of other options that you can play around with buried inside the code, feel free to explore !

Spline Fitting Toolbox

Version 0.11 [tgz]

Includes code for one dimensional cubic splines with and without derivative sampling, two dimensional thin plate splines with and without derivative sampling. Also included are two demo scripts which demonstrate the cubic and tps curve fitting procedures. The recent work of Dr. Jan Kybic on variational reconstruction forms the basis of most of this code. I would also like to thank Dr. Kybic for sharing his implementation of TPS curve fitting, which served as a benchmark for comparing the results of my own code.
History

09/13/2005 Version 0.11 (bugfix release) Small bugfix in tpsk.m due to Derek Johnson.
08/10/2002 Version 0.1 (initial release)

Spectral Clustering Toolbox

Spectral Clustering Toolbox version 0.2 [tgz]

Includes code for the original Normalized Cut algorithm (Shi & Malik) and the nystrom approximation (Fowlkes et al) . Multiway partitioning is implemented by performing k-means clustering in the eigenspace (Ng & Jordan).
History

07/24/2002 Version 0.2 Added code for performing color and proximity based image segmentation.
07/10/2002 Version 0.1 (Initial release)

Home