CacheTransform

Caches Transform::project() results.


CollectOutputTransform

DOCUMENT ME CHARLES


ContractTransform

It's like the opposite of ExpandTransform, but not really

Given a TemplateList as input, concatenate them into a single Template


CrossValidateTransform

Cross validate a trainable Transform.

Two flags can be put in File metadata that are related to cross-validation and are used to extend a testing gallery:

flag description
allPartitions This flag is intended to be used when comparing the performance of an untrainable algorithm (e.g. a COTS algorithm) against a trainable algorithm that was trained using cross-validation. All templates with the allPartitions flag will be compared against for every partition. As untrainable algorithms will have no use for the CrossValidateTransform, this flag is only meaningful at comparison time (but care has been taken so that one can train and enroll without issue if these Files are present in the used Gallery).
duplicatePartitions This flag is similar to allPartitions in that it causes the same template to be used during comparison for every partition. The difference is that duplicatePartitions will duplicate each marked template and project it into the model space constituded by the child transforms of CrossValidateTransform. Again, care has been take such that one can train with these templates in the used Gallery successfully (they will simply be omitted).

To use an extended Gallery, add an allPartitions="true" flag to the gallery sigset for those images that should be compared against for all testing partitions.


DirectStreamTransform

DOCUMENT ME CHARLES


DiscardTemplatesTransform

DOCUMENT ME


DiscardTransform

Removes all matrices from a Template.


DistributeTemplateTransform

DOCUMENT ME CHARLES


DownsampleTrainingTransform

DOCUMENT ME JOSH


EventTransform

DOCUMENT ME


ExpandTransform

Performs an expansion step on an input TemplateList. Each matrix in each input Template is expanded into its own Template.


FTETransform

Flags images that failed to enroll based on the specified Transform.


FirstTransform

Removes all but the first matrix from the Template.


ForkTransform

Transforms in parallel.

The source Template is seperately given to each transform and the results are appended together.


GalleryCompareTransform

Compare each Template to a fixed Gallery (with name = galleryName), using the specified distance. dst will contain a 1 by n vector of scores.


IdentityTransform

A no-op Transform.


IndependentTransform

Clones the Transform so that it can be applied independently.

Independent Transforms expect single-matrix Template.


JNITransform

Execute Java code from OpenBR using the JNI


LikelyTransform

Generic interface to Likely JIT compiler


LoadStoreDistance

Caches Distance training.


LoadStoreTransform

Caches Transform training.


PipeTransform

Transforms in series.

The source Template is given to the first transform and the resulting Template is passed to the next transform, etc.


ProcessWrapperTransform

Interface to a separate process


ProgressCounterTransform

DOCUMENT ME CHARLES


PropagateTransform

Ensures that a template will be propogated.


RemoveTransform

Removes the matrix from the Template at the specified index.


RestTransform

Removes the first matrix from the Template.


SchrodingerTransform

Generates two Template, one of which is passed through a Transform and the other is not. No cats were harmed in the making of this Transform.


SingletonTransform

A globally shared Transform.


StreamTransform

DOCUMENT ME CHARLES