An implementation of Cartesian k-means, M. Norouzi, D. J. Fleet, CVPR 2013.
After downloading the datasets, and compiling the mex files, RUN.m
will take you through training and testing of qunatization algorithms,
which are useful for approximate Euclidean nearest neighbor search.
See demo.m for a sample run on sift_1M dataset.
From within matlab please run the compile scripts (compile.m) in utils/ and search/ sub-directories to build the mex files.
Download the INRIA bigann datasets (two SIFT datasets and one GIST dataset) from http://corpus-texmex.irisa.fr/. Please make sure to modify RUN.m to point INRIA_HOME to the root folder of these datasets. INRIA_HOME folder should have the following sub-directories: matlab/, which includes the matlab I/O functions, and ANN_SIFT1M/, ANN_GIST1M/, and ANN_SIFT1B/, which include the training, and testing sets.
You can also download the Tiny images dataset (80 million GIST descriptors) from http://horatio.cs.nyu.edu/mit/tiny/data/index.html modify RUN.m to point INRIA_HOME to the root folder it. Unfortunately ground-truth nearest neighbor labels are not available for this dataset.
Copyright (c) 2013, Mohammad Norouzi <mohammad.n@gmail.com>. Please don't hesitate to drop me a line for bug reports or general comments. Thanks!
This is a free software; for license information please refer to license.txt file.