AlgorihthmEval: configuring over XML

The XML tag AlgorithmEval contains four elemens on top level:

        <LocateAlgorithm or RefineAlgorihtm or ProveAlgorihtm or AttentionAlgorithm />
        <AlgType></AlgType>
        <Eval></Eval>
        <AvgTime></AvgTime>

The first (necessarily the first) will be loaded by on of the plugin loader for Algorithms (Locate, Refine, Prove, Attention), this tag can have any name.

The second one specifies if the algorithm is able be iteratively called and if it can work to detect more than one object.

Eval and Avg time are the initial Evaluation.

Possible LocateAlgorithm are:

Name

Description

XML Configuration

Required Descriptors

FaceDetection

Uses OpenCV Face Detection to detect human faces in an RGB camera image

<FaceDetection/>

<DetectedFace/>

ShapeBased3DAlg

CAD matching

<ShapeBased3DAlg/>

<ShapeModel/>

TwoInOneAlg

Applies two Locatealgorithms sequentially

<TwoInOneAlg/>

None

IntersectTwoRuns

Applies a Locatealgorithms on two readings of different sensors.

<IntersectTwoRuns/>

None

CheckColorClass

Simple color classification

<CheckColorClass/>

<ColorClass/>

DeformShapeBasedAlg

Perspective shape model matching

<DeformShapeBasedAlg />

<DeformShapeModel/>

FindCalTab

Locate a Calbration plate

<FindCalTab/>

<CalTab/>

ClusterDetector

Extracts euclidean Clusters from range data

<ClusterDetector/>

<SegmentPrototype/>

!HClusterDetector

Extracts euclidean Clusters from range data

<HClusterDetector/>

<SegmentPrototype/>

SurfaceDetection

Matches 3D Surfaces

<SurfaceDetection/>

<SurfaceModel/>

DetectTransparentObjectCandidatet

Creates Candidates for Transparent objects

<DetectTransparentObjectCandidate />

<TransparentObjectCandidate/>

PanCakeDetector

Detects bright region inside a matched planar template (Names are more specialized than the method)

<PanCakeDetector />

<PanCakeMaker/>

DetectPlate

Locates the most complete circle of a specific size in an image (Names are more specialized than the method)

<DetectPlate />

<CircleDescriptor/>

Possible RefineAlgorithm are:

Name

Description

XML Configuration

Required Descriptors

Resulting Descriptor

RFAClassByDPs

Uses a descriptor based classification to determine a class based on a SVM classifier trained for a predefined set of classes

<RFAClassByDPs/>

<SegmentPrototype/>

<NamedClass/>

DetectTransparentObject

Uses a TransparentObjectCandidate to verify the transparancy at a certain location

<DetectTransparentObject/>

<TransparentObjectCandidate/>

<TransparentObject/>

RFADeformByCluster

Learns a planar shape model which can be matched perspective invariant

<RFADeformByCluster/>

<SegmentPrototype/>

<DeformShapeModel/>

RFAColorByShape

Learns a planar shape model which can be matched perspective invariant

<RFAColorByShape/>

<SegmentPrototype/>

<ColorClass/>

ShapeModelDownloader

Learns a planar shape model which can be matched perspective invariant

<ShapeModelDownloader/>

unbound class

<ShapeModel/>

CopOduRefine

Classifies objects using objects_of_daily_use_finder package

<CopOduRefine/>

<SegmentPrototype/>

<NamedClass/>

Possible Attention Algorithms are:

Name

Description

XML Configuration

Required Descriptors

Resulting Descriptor

FindBarCode

Reports all Barcodes in the current image, with an increasing certainty for repeated views.

<FindBarCode/>

none

<BarCode/>

Wiki: cop/AlgorithmEval (last edited 2011-05-16 15:56:56 by ThibaultKruse)