hCRF-light 2.0 (full version http://hcrf.sf.net)

This library is a light version of hCRF library (http://hcrf.sf.net) that contains implementations of HCRF[1] and LDCRF[2]. I added two new families of models, multiview counterparts of HCRF and LDCRF [3] and hierarchical sequence summarization approach of HCRF [4].

The source code is written in C++ using Visual Studio 2008. I tested the code on 32 bit and 64 bit versions of Windows 7 machines. I cannot guarantee its compatibility with other versions of Visual Studio. For your convenience, the library comes with precompiled excutables that work on a 32 bit and 64 bit Windows machines.

./hCRF-light/readme.md This file
./hCRF-light/apps Contains VS2008 project (hCRF.sln)
./hCRF-light/apps/matHCRF Matlab wrapper
./hCRF-light/apps/testMVLDM Command-line program
./hCRF-light/bin Contains command-line and matlab executables
./hCRF-light/bin/openMP Contains command-line and matlab executables (uses openMP)
./hCRF-light/libs/3rdParty Contains 3rd party libraries
./hCRF-light/libs/shared Contains hCRF library
./hCRF-light/matlab/ Matlab sample script
./hCRF-light/toydata/ Toy data in CSV & MAT format

Usage from command line (Windows 64)

1. Train and test HCRF with nbHiddenStates = 4
   .\bin\openMP\testModel64MP.exe -m hcrf -h 4 -g 0 -Fd .\toydata\dataTrain.csv -Fq .\toydata\seqLabelsTrain.csv -FD .\toydata\dataTest.csv -FQ .\toyData\seqLabelsTest.csv

2. Train and test HCNF with nbHiddenStates = 4, nbGates = 4
   .\bin\openMP\testModel64MP.exe -m hcrf -h 4 -g 4 -Fd .\toydata\dataTrain.csv -Fq .\toydata\seqLabelsTrain.csv -FD .\toydata\dataTest.csv -FQ .\toyData\seqLabelsTest.csv

3. Train and test HSS-HCRF with nbHiddenStates = 4, nbGates = 0, nbFeatureLayers = 4, segmentTau = 0.1
   .\bin\openMP\testModel64MP.exe -m hsshcrf -h 4 -g 0 -HL 4 -HT 0.1 -Fd .\toydata\dataTrain.csv -Fq .\toydata\seqLabelsTrain.csv -FD .\toydata\dataTest.csv -FQ .\toyData\seqLabelsTest.csv

4. Train and test HSS-HCNF with nbHiddenStates = 4, nbGates = 4, nbFeatureLayers = 4, segmentTau = 0.1
   .\bin\openMP\testModel64MP.exe -m hsshcrf -h 4 -g 4 -HL 4 -HT 0.1 -Fd .\toydata\dataTrain.csv -Fq .\toydata\seqLabelsTrain.csv -FD .\toydata\dataTest.csv -FQ .\toyData\seqLabelsTest.csv

See ./hCRF-light/matlab/test.m for examples in Matlab

Change log
ver 2.0 Added Hierarchical Sequence Summarization (HSS) HCRF [4]
Changed the name of the project [testMVLDM] to [testModels]
ver 1.1 Added a project matHCRF, a matlab wrapper
ver 1.0 Initial release

[1] Ariadna Quattoni, Sy Bor Wang, Louis-Philippe Morency, Michael Collins, Trevor Darrell: Hidden Conditional Random Fields. TPAMI 2007

[2] Louis-Philippe Morency, Ariadna Quattoni, Trevor Darrell: Latent-Dynamic Discriminative Models for Continuous Gesture Recognition. CVPR 2007

[3] Yale Song, Louis-Philippe Morency, Randall Davis: Multi-View Latent Variable Discriminative Models for Action Recognition. CVPR 2012

[4] Yale Song, Louis-Philippe Morency, Randall Davis: Action Recognition by Hierarchical Sequence Summarization. CVPR 2013