Plugins in version 4.4 of the CERT command line. Each of these plugins should have a plugin section in the config file that is distributed with CERT.

Face Detector
   Doesn't really detect faces, just collects the found face boxes and make them available to other plugins.

Blink Detector
   Obsolete. The SVM based "Blink AU" detector performs better.
   
Feature Detector Plugins
   Most of the feature detector plugins each detect one facial feature, e.g. left eye, right eye nasal (inner corner), etc. The eye detectors that are called "Left Eye Center" and "Right Eye Center", were both designed to be less effected by blinks than the other eye feature detectors.

Feature Improver Plugin
   The Feature Improver Plugin takes the outputs of ten feature detectors as input, and itself outputs eight 'improved' locations for those features. A linear regression model is used to improve the accuracy of the inputs.
   In order for the Feature Improver to work, all of the ten inputs features need to be enabled, although their outputs can be disabled. The ten input features are: Left Eye, Left Eye Temporal, Left Eye Nasal, Right Eye, Right Eye Temporal, Right Eye Nasal, Nose, Mouth, Mouth Left Corner and Mouth Right Corner. The eight outputs are improved versions of all of the inputs except the mouth corners, for which there was not enough training data.

Feature Averager
   Not a regular plugin, but it has a section of it's own in the 'global' section of the config file. There are only four feature 'slots' that are used to rotate and crop faces: left eye, right eye, nose and mouth. What the feature average does is allow you to average together several different features and use the average location in one of the cropping/rotation slots. The outer tag is called 'fdproc', inside that can be up to four other tags: 'left_eye', 'right_eye','nose' and 'mouth'. Inside each of those tags can be as many other feature id's (from the 'iname' tag in the feature's plugin section) as desired. A default setup should be included.

Gender Plugin
   Attempts to determine if a person is male or female. Higher values indicate a higher likelihood of a male face.

GKRPlugin
   The GKR plugin can do up to three post-processing steps on output from other plugins: GKR (spatial), logistic and CKF (temportal). The GKR processing requires a model of an individual's neutral face. The '-gt' and '-gc' parameters are used to train or use a GKR model.

GlassesPlugin
   Higher values indicate a higher liklihood that the person is wearing glasses.

IntervalCRP
   Calculates a likelihood that any interval contains a given AU (works on the base 11 AUs, which need to be enabled). Interval size is set in the config file. Results for a given frame represent the previous <interval> frames. Uses a Cross Recurrence Plot and Mean Time Trapping.

IntervalMean
   Calculates a likelihood that any interval contains a given AU (works on the base 11 AUs, which need to be enabled). Interval size is set in the config file. Results for a given frame represent the mean of the previous <interval> frames.

PosePlugin
   Outputs the yaw, pitch and roll values. Default output is in degrees but can be changed to radians in the config file.

SVMPlugin
   SVM plugins consist of an entry in the config file and a set of 72 weights files. Each set of weights files can detect multiple AUs each of which needs to have a label in that plugin's section of the config file. There are many examples of SVM plugins in the config file. The entry for an SVM plugin needs to point to the directory that contains it's weights files. The 'internal name' entry for an SVM plugin needs to match the prefix used to name all of the weights files.