/// <summary>
/// Train the EM model using the specific training data
/// </summary>
/// <param name="samples">The training data. A 32-bit floating-point, single-channel matrix, one vector per row</param>
/// <param name="sampleIdx">Can be null if not needed. When specified, identifies samples of interest. It is a Matrix>int< of nx1</param>
/// <param name="parameters">The parameters for EM</param>
/// <param name="labels">Can be null if not needed. Optionally computed output "class label" for each sample</param>
/// <returns></returns>
public bool Train(Matrix<float> samples, Matrix<float> sampleIdx, EMParams parameters, Matrix<Int32> labels)
{
MCvEMParams param = new MCvEMParams();
param.nclusters = parameters.Nclusters;
param.cov_mat_type = parameters.CovMatType;
param.start_step = parameters.StartStep;
param.probs = parameters.Probs == null ? IntPtr.Zero : parameters.Probs.Ptr;
param.means = parameters.Means == null ? IntPtr.Zero : parameters.Means.Ptr;
param.weights = parameters.Weights == null ? IntPtr.Zero : parameters.Weights.Ptr;
GCHandle? covsPtrHandle = null;
if (parameters.Covs == null)
param.covs = IntPtr.Zero;
else
{
IntPtr[] covsPtr = Array.ConvertAll<CvArray<double>, IntPtr>(parameters.Covs,
delegate(CvArray<double> m) { return m.Ptr; });
covsPtrHandle = GCHandle.Alloc(covsPtr, GCHandleType.Pinned);
param.covs = covsPtrHandle.Value.AddrOfPinnedObject();
}
param.term_crit = parameters.TermCrit;
bool res = MlInvoke.CvEMTrain(
_ptr,
samples.Ptr,
sampleIdx == null? IntPtr.Zero : sampleIdx.Ptr,
param,
labels == null? IntPtr.Zero : labels.Ptr);
if (covsPtrHandle.HasValue)
covsPtrHandle.Value.Free();
return res;
}
/// <summary>
/// Train the EM model using the specific training data
/// </summary>
/// <param name="samples">The training data. A 32-bit floating-point, single-channel matrix, one vector per row</param>
/// <param name="parameters">The parameters for EM</param>
/// <param name="labels">Can be null if not needed. Optionally computed output "class label" for each sample</param>
/// <returns></returns>
public bool Train(Matrix <float> samples, EMParams parameters, Matrix <Int32> labels)
{
MCvEMParams param = new MCvEMParams();
param.nclusters = parameters.Nclusters;
param.cov_mat_type = parameters.CovMatType;
param.start_step = parameters.StartStep;
param.probs = parameters.Probs == null ? IntPtr.Zero : parameters.Probs.Ptr;
param.means = parameters.Means == null ? IntPtr.Zero : parameters.Means.Ptr;
param.weights = parameters.Weights == null ? IntPtr.Zero : parameters.Weights.Ptr;
GCHandle?covsPtrHandle = null;
if (parameters.Covs == null)
{
param.covs = IntPtr.Zero;
}
else
{
IntPtr[] covsPtr =
#if NETFX_CORE
Extensions.
#else
Array.
#endif
ConvertAll <Matrix <double>, IntPtr>(parameters.Covs, delegate(Matrix <double> m) { return(m.Ptr); });
covsPtrHandle = GCHandle.Alloc(covsPtr, GCHandleType.Pinned);
param.covs = covsPtrHandle.Value.AddrOfPinnedObject();
}
param.term_crit = parameters.TermCrit;
bool res = MlInvoke.CvEMLegacyTrain(
_ptr,
samples.Ptr,
IntPtr.Zero,
param,
labels == null ? IntPtr.Zero : labels.Ptr);
if (covsPtrHandle.HasValue)
{
covsPtrHandle.Value.Free();
}
return(res);
}
/// <summary>
/// Creaet an Expectation Maximization model using the specific training parameters
/// </summary>
/// <param name="samples">The training data. A 32-bit floating-point, single-channel matrix, one vector per row</param>
/// <param name="parameters">The parameters for EM</param>
/// <param name="labels">Can be null if not needed. Optionally computed output "class label" for each sample</param>
public EMLegacy(Matrix <float> samples, EMParams parameters, Matrix <Int32> labels)
: this()
{
Train(samples, parameters, labels);
}
/// <summary>
/// Creaet an Expectation Maximization model using the specific training parameters
/// </summary>
/// <param name="samples">The samples to be trained</param>
/// <param name="sampleIdx"></param>
/// <param name="parameters"></param>
/// <param name="labels"></param>
public EM(Matrix<float> samples, Matrix<float> sampleIdx, EMParams parameters, Matrix<Int32> labels)
: this()
{
Train(samples, sampleIdx, parameters, labels);
}
/// <summary>
/// Creaet an Expectation Maximization model using the specific training parameters
/// </summary>
/// <param name="samples">The samples to be trained</param>
/// <param name="sampleIdx"></param>
/// <param name="parameters"></param>
/// <param name="labels"></param>
public EM(Matrix <float> samples, Matrix <float> sampleIdx, EMParams parameters, Matrix <Int32> labels)
: this()
{
Train(samples, sampleIdx, parameters, labels);
}
