Home

Proximity Forest Wiki

Welcome to the wiki pages for this project. An overview of the Proximity Forest for computing approximate nearest neighbors can be found in the paper "Are You Using the Right Approximate Nearest Neighbor Algorithm?", by S. O'Hara and B.A. Draper, Workshop on Applications of Computer Vision (WACV), 2013. It earned a best student paper award.

This project also contains the implementation of Subspace Forests for fast, accurate, and scalable action recognition in video. See the paper: "Scalable Action Recognition using a Subspace Forest" by S. O'Hara and B.A. Draper, IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2012. A Subspace Forest is a type of Proximity Forest where the samples are subspaces and the distance function is based on the canonical angles between subspaces.

The current implementation is in Python 2.7.

The Proximity Forest uses the PyVision open source computer vision library for Python (also hosted on SourceForge, and to which I contribute), but this is mainly only needed for the Subspace Forest action recognition code and some demonstrations. If you are just using the Proximity Forest, you probably don't need pyvision. Let me know if you have any problems in this regard. More information on installation and requirements can be found on this wiki's installation page.

[Installation]
[CVPR2012_Experiments]
[WACV2013_Experiments]

Project Admins:

• Stephen O'Hara