This code was used to produce the results described in the paper:
"Recognizing human actions in still images: a study of bag-of-features and 
part-based representations", Vincent Delaitre, Ivan Laptev, Josef Sivic, In
Proceedings of the British Machine Vision Conference.

If you use this code, please cite our work:

@inproceedings{Delaitre10,
  title = "Recognizing human actions in still images: a study of bag-of-features and part-based representations",
  author = "Delaitre, V. and Laptev, I. and Sivic, J.",
  year = "2010",
  address = "Aberystwyth, United Kingdom",
  booktitle = "Proceedings of the British Machine Vision Conference",
}


Quick start
===========

Change Matlab's current directory into the directory of this README and 
setup Matlab's path by typing "bmvc_setup" in Matlab command line.

You can launch our code by running :

>> run_classifier('DB/Willow-actions', {'trainval' 'test'}, 'C2'); 

The first parameter is the path to the database you want to use. It available on
http://www.di.ens.fr/willow/research/stillactions/willowactions.zip

The second argument defines the training set and the testing set. Willow-actions
uses the same conventions that the PASCAL VOC challenge uses. The training set
is made of two sub-sets: the training and validation sets, referenced by 'train'
and 'val' respectively. The concatenation of the two is designated by 
'trainval'. The testing set is called 'test'.
The third argument is the configuration you want to use: 'A', 'B', 'C1' or 'C2'.
See the paper for more informations about configurations.

See 'examples.m' for more detailed examples about 'run_classifier'.