VisionLab Features Library
VLFeat is a collection of popular computer vision algorithms (such as SIFT and MSER). Algorithms are implemented in a clean C library and there are command line and MATLAB based interfaces. VLFeat also includes many useful MATLAB functions. VLFeat is distributed under GNU GPL.Please note that VLFeat is still unstable.
SIFT, SIFT++, MSER, and VLUtil
SIFT is an implementation of D. Lowe's Scale Invariant Feature Transform for MATLAB. It is quite faithful to the original and easy to modify. SIFT++ is a command-line version which runs much faster. MSER is an implementation of Maximally Stable Extremal Regions in MATLAB, but written in C and relatively efficient. VLUtil is a collection of other MATLAB programs. Some are quite handy and we use them in our workflow. They are also used for some of the demos contained in the other packages.This code is stable, but it will be made obsolete by VLFeat.
Bag of Features
Bag is an implementation fo a simple Bag of Feature classifier for visual recognition. The method is tested on Caltech-4 and Caltech-101.Others
MDoc is a Perl
script that exports the documentation embedded
in MATLAB M-Files to an HTML page.
Autorights is
a Perl script that adds a copyright notice to
several source files at once.
VLPov contains
MATLAB functions and a patch to MegaPOV to
generate virtual datasets useful for computer
vision applications.
Demos of
famus algorithms (e.g. AdaBoost and
Blatuh-Arimoto) and toy programs. Educational
value.
Anaview is a
MATLAB M-file that converts 3-D plots into
anaglyps.