[Imageworld] software to extract classemes

[lorenzo at cs.dartmouth.edu]

We are pleased to announce the release of our software to extract the classeme
descriptor introduced in:
Efficient Object Category Recognition Using Classemes, Lorenzo Torresani, Martin
Szummer and Andrew Fitzgibbon, ECCV 2010.

The descriptor is very compact (333 bytes/image for the binary version) but it
provides near state-of-the-art object categorization accuracy even with simple
linear classifiers. The average extraction time is less than 2 seconds per
image.

The software is available for Microsoft Windows, GNU/Linux and Mac OSX.

See the following website for code and description:
http://vlg.cs.dartmouth.edu/projects/classemes_extractor/

Lorenzo Torresani