Section: New Software and Platforms

Efficient video descriptors for action recognition

This package contains source code for highly-efficient extraction of local space-time video descriptors for action recognition. The accuracy of descriptors measured at standard benchmarks for action recognition is comparable to the state-of-the-art dense trajectory features, while being more than 100 times faster on standard CPU. The previous version of this code was evaluated in our recent work [12] . The package is available from http://www.di.ens.fr/~laptev/download/fastvideofeat-1.0.zip . Earlier version of our space-time video features is available at http://www.di.ens.fr/~laptev/download/stip-2.0-linux.zip .