vlfeat/vlfeat

Please refer to the project homepage (http://www.vlfeat.org) for software releases, binary packages, documentation, examples, and tutorials.
VLFeat is a popular library of computer vision algorithms with a focus on local features (SIFT, LIOP, Harris Affine, MSER, etc) and image understanding (HOG, Fisher Vectors, VLAD, large scale discriminative learning).

VLFeat is used in research for fast prototyping, as well as in education as the basis of several computer vision laboratories. It is fully integrated in MATLAB, but provides a C API as well.

VLFeat has been under development since 2007 and has been cited in more than 950 scientific publications.

VLFeat is authored by a team of computer vision researchers at Oxford, UCLA, and several other institutions as well as individual contributors. It is released under the BSD license.