TCD (v1.3)

This is a toy demo for the unsupervised Temporal Commonality Discovery (TCD) using a branch and bound approach.

Created: Oct-6-2012 by Wen-Sheng Chu (wschu@cmu.edu)
Updated: Aug-8-2014 by Wen-Sheng Chu (wschu@cmu.edu)
Citation: Unsupervised Temporal Commonality Discovery, ECCV, 2012.

and/or cite the bibtex

@inproceedings{chu2012unsupervised,
    title={Unsupervised temporal commonality discovery},
    author={Chu, Wen-Sheng and Zhou, Feng and {De la Torre}, Fernando},
    booktitle={ECCV},
    year={2012}
}

Contents

The project tree is organized as follows.

lib/            directory for utility functions
src/            directory for our c++ implementation of TCD algo
README.md       this file
demoTCD.m       demo of TCD on discovering temporal commonalities among two sequences 

Usage

In Matlab shell:

>> cd tcd_v1.3
>> make
>> addpaths
>> demoTCD

You may want to change pickEx in the script to play with other examples.

The script has been tested on Windows 8, Ubuntu 3.11.0 and Mac X 10.9.4 without problems. Please send your feedbacks to Wen-Sheng Chu regarding any issues, bugs and improvements.

Change Log

Aug-08-2014

*    Implemented the main TCD routine in C++
*    The speedup against v1.2 is fascinating and can be found [here](http://humansensing.cs.cmu.edu/wschu/project_tcd.html#FIG_SPEEDUP)

Jan-06-2013

*    Extend TCD with soft clustering and hard clustering
*    Speedup TCD search using integral image
*    Speedup multiple TCD search using queue elimination

Oct-06-2012

*    Add the main implementation of the TCD algorithm
*    Add comparisons to the naive SW approach
*    Add a demo for 8 synthetic examples, including detecting multiple commonalities and an application to video indexing

License

The code may be redistributed under BSD license. Please send your feedbacks to Wen-Sheng Chu. :)