fex-metrica comprises a set of preprocessing and statistical tools for the analysis of time series of facial expressions. This toolbox contains Matlab functions and classes for preprocessing the time series computed with FACET SDK and Emotient Analytics facial expressions recognition tools developped by Emotient, Inc.
doc/ documnetation [not developped]
fexSDK/ software development kit
samples/ directory for sample application codes
shared/ directory with shared hardcoded information
src/ source code directory
facet/ directory with cpp file for FacetSDK
fexc.m master class for fex-metrica
ui/ directory for user interface
util/ directory for utilities
viewer/ visualization toolbox
test/ directory for tests
fexinstall.m installation script
LICENSE.md MIT license
README.md this file
Analytic tools can be access using fexc.m, which defines the main object used in fex-metrica. This class wraps the Emotient, Inc toolbox for video analysis, as well as generic video manipulation utilities, and timeseries transformations. The tools associated with a FEXC object include:
- facet: a set of hard-coded routines which wrap some of FACET SDK functions;
- util: several functions for data handling, and time series manipulation;
- ui: A user interface [partially developped].
- viewer: toolbox for visualization.
Fex-metrica was developped on Unix. Most operation should work on Windows as well, but they were never tested.
The following Matlab modules are required:
- Matlab stats toolbox;
- Matlab computer vision toolbox;
- Matlab signal processign toolbox.
Additionally, the VIEWER functions require ffmpeg (or avconv).
When using Emotient Analytics, the facet module is not required. Instruction for installation using Facet SDK are included in the facet directory.
Start Matlab, and navigate to the main fex-metrica directory. On the Matlab prompt, type the following:
>> fexinstall(1)
This will:
- Run some tests;
- Generate a fex_init.m file;
- Unzip the example folder.
After installation, in order to initialize fex-metrica type:
fex_init;
The doc folder is empty. For now, documentation for most functions or methods can be accessed from Matlab, using "help" or "doc."
The samples folder contains examples of some operations that you can carry on in fex-metrica.
Method | Update | isDone? |
---|---|---|
.FEXC | Update documentation | 0 |
Fix timing issue | 0 | |
.UPDATE | Add multiple args | 0 |
.DESIGN & .DESIGNINIT | 0 | |
.REINITIALIZE | Use defaults | 0 |
.FEXPORT | Clean code | 0 |
Remove method "data1" | 0 | |
.DERIVESENTIMENTS | Make Sentiments a getter func | 0 |
Set .thrsemo as structure | 0 | |
Add two thresold method | 0 | |
.DOWNSAMPLE | Fix structural & video | 0 |
Gaussian kernel option | 0 | |
.SETBASELINE | Make a private property | 0 |
Improve "neutral options" | 0 | |
.INTERPOLATE | Matrix size with structural | 0 |
.GETMATRIX | Implement as MATRIX | 0 |
.GETBAND | Implement | 0 |
.NORMALIZE | Add arbitrary bounds | 0 |
.FEXEXPORT2VIEWER | Remove | 0 |
.SHOWANNOTATION | Make stakable in GET | 0 |
.REGRESS | Implement MLR | 0 |
.CLASSIFY | Implement MLR | 0 |
.SHOWANNOTATION | Make stakable in GET | 0 |
.DEFAULT | Global default files | 0 |
.ESTCRF | Canonical Resp. Project | 0 |
.TEST | Implement tests | 0 |
===========
Support for this research was provided by NSF SBE 1232676. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of NSF.