Basic Random Forest implementation for Matlab ©

Short description

Implementation of Random Forest learning and classification with optional out-of-bag error estimation but without any other fancy features.

Download

July 2013 release: [ ZIP ]

Active Contour Toolbox for Matlab ©

Short description

The toolbox provides some functions for deforming polygons in order to implement image or video segmentation by means of active (or deformable) contours. Contour topology is handled in a way that should allow changes similar to what can be observed with level sets (merging, splitting, and hole creation by selfmerging).

Illustration & Download

Active Contour Toolbox Illustration


Download — May 2010 release: [ ZIP ]

Nearest neighbor (kNN) Search with CUDA™

Short description

Given a vector x in n dimensions and many vectors Y = {y_i, i=1..q}, the k-nearest neighbor (kNN) search consists in extracting and ordering the k vectors in Y closest to x. In practice, one often needs to perform this search not only for a singleton x but for each element of a set {x_i, i=1..p}. This is very time consuming. On the other hand, such a task is twofold-parallelizable: each search can be performed independently of the others, and for a given search, the main calculations are independent of each other. Graphics cards are typically built to perform many (graphical) operations in parallel. This processing power can be exploited for scientific computing. The “kNN search using GPU” is a code developed for CUDA™-enabled graphics cards to perform kNN searches efficiently.

Download