kd-GeRaF

kd-GeRaF is a C++ header-only library, whcich provides a Generalized Randomized k-d Forest for (approximate) Nearest Neighbor Search in a high dimensional space.

How to use kd-GeRaF?

Just include kd-GeRaF’s header file.

How fast is kd-GeRaF??

We handle GIST datasets of 10⁶ images in 960 dimensions in < 1 sec with about 90% outputs being true nearest neighbors.

Project/thesis of George Samaras, university of Athens, 2015.

• Foler ‘Doxygen’ contains the html documentation.
• Folder ‘main’ contains files that can be the main() of the project.
• Folder ‘makefiles’ contains several Makefiles.
• Folder ‘source’ contains the source code of kd-GeRaF.
• Folder ‘test_files’ contains files used by the files in ‘main’ folder.
• Makefile should be used for compilation.
• LICENSE.txt contains the license of kd-GeRaF.
• run.sh contains a script for running experiments.

Read more in our Paper: High-dimensional approximate nearest neighbor: k-d Generalized Randomized Forests[Avrithis, Prof. Emiris and Samaras].