### Installation and Requirements###

You should be able to run the code with MATLAB R2010b or later. 
The package includes one section of my own toolbox, if you want you
could download the full version from the following github website: 

    http://github.com/lbl1985/toolbox

The code uses tracklets obtained using code provided by [2]. You can obtain
the code at

    http://lear.inrialpes.fr/people/wang/dense_trajectories

Our package contains two versions of binary files compiled from
their source code (one with velocities and the other with absolute coordinates).

We are working on a CUDA implementation that will be uploaded soon. Please 
check again in a few weeks at the repo:

    http://github.com/lbl1985/hankelet

### Tests ###

The testing script mainProcedure.m has two sections. One is for
massive processing and the other is for labeling visualization.

Run the following code in MATLAB terminal: 

>> mainProcedure

---- Test  Section 1 ---- 
This section loads the testData.mat, which is velocity information
extracted from tracklets from a set of videos depicting activities. 

This section includes Kmeans Clustering and Labeling
subsections. These two steps are the basis for Bags of Hankelets as 
described in [1].


Based on these procedures, other researchers can easily modify the code to use
their own experiment settings for other datasets. 

---- Test  Section 2 ---- 
This section loads person_trial.mat, which is extracted from
person01_boxing_d1_uncomp video in KTH dataset [3] by using the two binary
files located in 90support/dense_binary.  Again, one is for velocities and 
the other one is for absolute coordinates.

The variable X_train is extracted from "corresponding velocity" (:, 8:37)';
and varialbe trajectory is extracted from "absolute" (:, 8:37);

The "corresponding velocity" is the information we processed in
our algorithm for clustering and labeling. The absolute coordinates portion
 is only for visulaization purposes. 

References:

[1] B.Li and O. I. Camps, M. Sznaier, "Cross-view Activity Recognition using
Hankelets," presented at the Computer Vision and Pattern Recognition
(CVPR), 2012 IEEE Conference on, 2012, pp. 1362-1369

[2] H.Wang, A. Klaser, C. Schmid, and C.-L. Liu, "Action
recognition by dense trajectories," presented at the Computer Vision
and Pattern Recognition (CVPR), 2011 IEEE Conference on, 2011,
pp. 3169-3176.

[3] C. Schuldt, I. Laptev, B. Caputo, "Recognizing Human Actions: A Local
SVM Approach," presented at the International Conference on Pattern Recognition
(ICPR), 2004, IEEE Conference on, 2012, pp. 32-36

### Bugs and extensions ###

If you find bugs, etc., feel free to drop me a line. Also if you
developed some extension to the program, let me know and I can include
it in the code. You could contact me through email: 

binlong.li at gmail.com

### Citation ###

If you use this code, please cite our paper [1]:

@inproceedings{LiCaSzCVPR12,
author={B. Li and O. I. Camps and M. Sznaier},
title={{Cross-view Activity Recognition using
Hankelets}},
booktitle={{IEEE} Conference on Computer Vision \& Pattern Recognition},
year= {2012},
month={June},
address={Providence, Rhode Island},
pages={1362--1369}}


### LICENSE CONDITIONS ###

Copyright (C) 2012 Binlong Li, Octavia I Camps, Mario Sznaier

This program is free software; you can redistribute it and/or
modify it under the terms of the GNU General Public License
as published by the Free Software Foundation; either version 2
of the License, or (at your option) any later version.

This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
GNU General Public License for more details.

You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA  02110-1301, USA.