Updated: 19/06/2014 Previous updates: 25/10/2013 VERSION 1.0 =============================================================================== Matlab Software for: - FLAT MM (Flat Markov Model) - CLHBC (Cartesian product Label based Hierarchical Bottom-up Clustering) - MLCTP (Multi-level Chinese Takeaway Process) - Two Hybrids (MLCTP-BW and MLCTP-CLHBC) =============================================================================== Copyright (C) 2013, Aftab Khan and David Windridge (aftabkhan00[at]gmail[dot]com and d.windridge[at]surrey[dot]ac[dot]uk) This software package includes several Matlab scripts and auxiliary functions/utilities, which implement rule induction algorithms in the context of various domains (sports and non-sports). Models here are described in the following publication: Multi-Level Chinese Takeaway Process and Label-Based Processes for Rule Induction in the Context of Automated Sports Video Annotation Aftab Khan, David Windridge, Josef Kittler IEEE Transactions on Cybernetics, 2014 Please cite this paper in any publications using the MLCTP package. ======================================================================== Package Organization and Documentation ======================================================================== To generate topologies for MLCTP - go to MLCTP_Topology_creation folder and run MLCTP_Topology_Generator (specify alpha and gamma inside that code (or the range of these hyper-parameters - It will generate a mat file with a collection of topologies which either needs copying over to this folder or adding MLCTP_Topology_creation folder as addpath('./MLCTP_Topology_Generator'). Source footage files are not present in this folder due to licensing restrictions. However, annotated features are present in the dataset folder that could be used in conjunction with this code. In principle any sequential data can be presented to this code for rule induction. This may require minor tweaks to be made to the code. In the next version, I hope to make this code more generally usable on all types of sequential data. Activity and MSNBC.com source datasets are available at the UCI repository; It's filtered version is in the dataset folder. Example run: Generate_Results(5,1) %for Activity dataset with Flat MM Generate_Results(1,2) %for Badminton using CLHBC Generate_Results(2,3) %for Tennis using MLCTP (Note, this may require generating the topologies first in "MLCTP_Topology_creation" folder (file names must be matched) %Note, source code for running the hybrid models is available in the "Hybrids" folder This package may require additional resources to run properly. ======================================================================== Acknowledgements ======================================================================== This package was developed as part of the ACASVA project (http://www.cvssp.org/acasva) with contributions/suggestions from: Dr. Teofilo de Campos Dr. William Christmas Dr. Fei Yan Dr. Ibrahim Almajai Miss. Nazli FarajiDavar Prof. Josef Kittler ======================================================================== Copyright & License ======================================================================== Copyright (C) 2014, Aftab Khan and David Windridge. http://www.cvssp.org/acasva Permission is granted for anyone to copy, use, or modify these programs and accompanying documents for purposes of research or education, provided this copyright notice is retained, and note is made of any changes that have been made. These programs and documents are distributed without any warranty, express or implied. As the programs were written for research purposes only, they have not been tested to the degree that would be advisable in any important application. All use of these programs is entirely at the user's own risk.