Class GPCForest5

Namespace

OpenCvSharp.OptFlow

Assembly

OpenCvSharp.dll

Class for computing the correspondences between two images using the Global Patch Collider (GPC) algorithm.

public class GPCForest5 : Algorithm, IDisposable

Inheritance

object

CvObject

CvPtrObject

Algorithm

GPCForest5

Implements

IDisposable

Inherited Members

Algorithm.Write(FileStorage)

Algorithm.Read(FileNode)

Algorithm.Empty

Algorithm.Save(string)

Algorithm.GetDefaultName()

CvPtrObject.RawPtr

CvObject.IsDisposed

CvObject.ptr

CvObject.SetSafeHandle(OpenCvSafeHandle)

CvObject.Dispose()

CvObject.Dispose(bool)

CvObject.DisposeManaged()

CvObject.DisposeUnmanaged()

CvObject.ThrowIfDisposed()

CvObject.CvPtr

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

This wraps cv::optflow::GPCForest<T>, a C++ template class, with a single fixed explicit instantiation at T = 5 (matching what upstream OpenCV's own samples/tests universally hardcode).

Methods

Create()

Creates an empty GPC forest of 5 trees.

public static GPCForest5 Create()

Returns

GPCForest5

FindCorrespondences(InputArray, InputArray, out Point[], out Point[], GPCMatchingParams?)

Finds correspondences between two images using the trained forest.

public void FindCorrespondences(InputArray imgFrom, InputArray imgTo, out Point[] pointsFrom, out Point[] pointsTo, GPCMatchingParams? matchingParams = null)

Parameters

imgFrom InputArray

First image in a sequence (3-channel).

imgTo InputArray

Second image in a sequence (3-channel).

pointsFrom Point[]

Coordinates in imgFrom of each found correspondence.

pointsTo Point[]

Corresponding coordinates in imgTo, parallel to pointsFrom.

matchingParams GPCMatchingParams

Additional matching parameters for fine-tuning. Uses the native defaults when null.

Train(GPCTrainingSamples, GPCTrainingParams?)

Trains the forest using one sample set for every tree. Consider using the overload taking image/ground-truth-flow triples instead for better quality.

public void Train(GPCTrainingSamples samples, GPCTrainingParams? trainingParams = null)

Parameters

samples GPCTrainingSamples

Training samples, e.g. obtained via Create(IEnumerable<Mat>, IEnumerable<Mat>, IEnumerable<Mat>, GPCDescType).

trainingParams GPCTrainingParams

Training parameters. Uses the native defaults when null.

Train(IEnumerable<Mat>, IEnumerable<Mat>, IEnumerable<Mat>, GPCTrainingParams?)

Trains the forest using individual samples extracted for each tree from the given image pairs and ground-truth flow. Generally produces better quality than the single-sample-set overload.

public void Train(IEnumerable<Mat> imagesFrom, IEnumerable<Mat> imagesTo, IEnumerable<Mat> gt, GPCTrainingParams? trainingParams = null)

Parameters

imagesFrom IEnumerable<Mat>

First images of the image pairs.

imagesTo IEnumerable<Mat>

Second images of the image pairs.

gt IEnumerable<Mat>

Ground-truth flow for each image pair.

trainingParams GPCTrainingParams

Training parameters. Uses the native defaults when null.