ASEFEyesTransform

Find eye locations using an ASEF filter

  • file: metadata/eyes.cpp
  • inherits: UntrainableTransform
  • author(s): Josh Klontz

  • read:

    1. Bolme, D.S.; Draper, B.A.; Beveridge, J.R.;
      "Average of Synthetic Exact Filters," Computer Vision and Pattern Recognition, 2009. CVPR 2009. IEEE Conference on , vol., no., pp.2105-2112, 20-25 June 2009

  • properties: None

AnonymizeLandmarksTransform

Remove a name from a point/rect

AsTransform

Change the Template file extension

AveragePointsTransform

Averages a set of landmarks into a new landmark

CascadeTransform

Wraps OpenCV cascade classifier

CheckPointsTransform

Checks the points in a template for missing (-1,-1) values

CheckTransform

Checks the Template for NaN values.

ClearPointsTransform

Clears the points from a Template

ConsolidateDetectionsTransform

Consolidate redundant/overlapping detections.

CropRectTransform

Crops the width and height of a Template rects by input width and height factors.

DelaunayTransform

Creates a Delaunay triangulation based on a set of points

ExpandRectTransform

Expand the width and height of a Template rects by input width and height factors.

ExtractMetadataTransform

Create matrix from metadata values.

FaceFromEyesTransform

Create face bounding box from two eye locations.

  • file: metadata/facefromeyes.cpp
  • inherits: UntrainableMetadataTransform
  • author(s): Brendan Klare

  • properties:

    Property Type Description
    widthPadding double Specifies what percentage of the interpupliary distance (ipd) will be padded in both horizontal directions.
    verticalLocation double specifies where vertically the eyes are within the bounding box (0.5 would be the center).

FileExclusionTransform

DOCUMENT ME CHARLES

FilterDupeMetadataTransform

Removes duplicate Template based on a unique metadata key

FilterMetadataTransform

Filters templates such that the only remaining template wills have metadata values witihin the the specified ranges.

  • file: metadata/filtermetadata.cpp
  • inherits: UntrainableMetaTransform
  • author(s): Brendan Klare

  • properties:

    Property Type Description
    key1 QString The meta-data key(s) to filter on
    value1 float The values to compare the values of key1 and key2 entires against
    compareType1 QString The comparison operation to perform. "le" -> val(key1) <= value1, "lt" -> val(key1) < value1,
    "ge" -> val(key1) >= value1, "gt" -> val(key1) > value1, "eq" -> val(key) == value1.

GridTransform

Add landmarks to the Template in a grid layout

GroundTruthTransform

Add any ground truth to the Template using the file's base name.

HOGPersonDetectorTransform

Detects objects with OpenCV's built-in HOG detection.

IPDTransform

Calculate interpupillary distance or the distance between two arbitrary points.

  • file: metadata/ipd.cpp
  • inherits: UntrainableMetadataTransform
  • author(s): [Ben Klein][bhklein]

  • properties:

    Property Type Description
    named bool Are the points named?
    firstEye QString First point's metadata key.
    secondEye QString Second point's metadata key.
    key QString Metadata key for distance.
    indices QList Indices of points in metadata if not named.

IfMetadataTransform

Clear Template without the required metadata.

ImpostorUniquenessMeasureTransform

Impostor Uniqueness Measure

  • file: metadata/imposteruniquenessmeasure.cpp
  • inherits: Transform
  • author(s): Josh Klontz
  • properties: None

JSONTransform

Represent the metadata as JSON template data.

JustTransform

Preserve the contents of the input template, updating just the specified metadata keys.

KeepMetadataTransform

Retains only the values for the keys listed, to reduce template size

KeyPointDetectorTransform

Wraps OpenCV Key Point Detector

KeyToRectTransform

Convert values of key_X, key_Y, key_Width, key_Height to a rect.

NameLandmarksTransform

Name a point/rect

NameTransform

Sets the Template matrix data to the br::File::name.

NormalizeMetadataTransform

Normalize a floating point metadata item to be between -1 and 1. Useful for classifier labels.

  • file: metadata/normalizemetadata.cpp
  • inherits: UntrainableMetadataTransform
  • author(s): Scott Klum

  • properties:

    Property Type Description
    inputVariable QString Metadata key for the item to scale. Default is empty string.
    min float Minimum possible value for the metadata item (will be scaled to -1). Default is -1.
    min float Maximum possible value for the metadata item (will be scaled to 1). Default is 1.

NormalizePointsToRectTransform

Scales a floating point metadata item by index

  • file: metadata/normalizepointstorect.cpp
  • inherits: UntrainableTransform
  • author(s): Scott Klum

  • properties:

    Property Type Description
    index int Index of rect to normalize points with. If negative, will normalize the points to the width of the matrix. Default is -1.

NormalizePointsTransform

Normalize points to be relative to a single point

PointDisplacementTransform

Normalize points to be relative to a single point

PointsToMatrixTransform

Converts either the file::points() list or a QList metadata item to be the template's matrix

ProcrustesTransform

Procrustes alignment of points

RectToKeysTransform

Convert a rect to X, Y, Width, and Height. Handy for saving universal templates.

RectsToMatsTransform

For each rectangle bounding box in src, the mat is cropped and appended the the template's list of mats.

RectsToTemplatesTransform

For each rectangle bounding box in src, a new Template is created.

RegexPropertyTransform

Apply the input regular expression to the value of inputProperty, store the matched portion in outputProperty.

RemoveMetadataTransform

Removes a metadata field from all Template

RemoveTemplatesTransform

Remove Template with the specified file extension or metadata value.

RenameFirstTransform

Rename first found metadata key

RenameTransform

Rename metadata key

ReorderPointsTransform

Reorder the points such that points[from[i]] becomes points[to[i]] and vice versa

RestoreMatTransform

Set the last matrix of the input Template to a matrix stored as metadata with input propName.

Also removes the property from the Templates metadata after restoring it.

SaveMatTransform

Store the last matrix of the input Template as a metadata key with input property name.

ScaleMetadataTransform

Scales a floating point metadata item by factor

  • file: metadata/scalemetadata.cpp
  • inherits: UntrainableMetadataTransform
  • author(s): Scott Klum

  • properties:

    Property Type Description
    inputVariable QString Metadata key for the item to scale. Default is empty string.
    factor float Floating point factor to scale the item. Default is 1.

ScaleRectsTransform

Scales a rectangle either by pixel value or factor.

  • file: metadata/scalerects.cpp
  • inherits: UntrainableTransform
  • author(s): Scott Klum

  • properties:

    Property Type Description
    index int Index of rect to scale.
    left int Number of pixels to scale the left of the rect.
    top int Number of pixels to scale the top of the rect.
    right int Number of pixels to scale the right of the rect.
    bottom int Number of pixels to scale the bottom of the rect.
    factor float Scale factor.

SelectPointsTransform

Retains only landmarks/points at the provided indices

SetMetadataTransform

Sets the metadata key/value pair.

SetPointsInRectTransform

Set points relative to a rect

SquareFromPointsTransform

Creates a bounding square around three points (typically the two eyes and the chin).

  • file: metadata/squarefrompoints.cpp
  • inherits: UntrainableMetadataTransform
  • author(s): Scott Klum

  • properties:

    Property Type Description
    eyeL int index of left eye point.
    eyeR int Index of right eye point.
    chin int Index of chin point.

StasmTransform

Wraps STASM key point detector

StopWatchTransform

Gives time elapsed over a specified Transform as a function of both images (or frames) and pixels.

  • file: metadata/stopwatch.cpp
  • inherits: MetaTransform
  • author(s): [Jordan Cheney][JordanCheney], Josh Klontz
  • properties: None

VerifyDetectionTransform

Check if a the automated face detection contains the landmarks correpsonding to the tempate metadata. If not, drop the template. This is meant for use during training, where the landmarks will be ground truth'd. If one wants to using a ground truth bounding box instead, then convert the BB to a landmark.