46 lines
1.3 KiB
Go
46 lines
1.3 KiB
Go
// Copyright 2014 <t.kastner@cumulo.at>. All rights reserved.
|
|
// Use of this source code is governed by a BSD-style
|
|
// license that can be found in the LICENSE file.
|
|
|
|
package opencv
|
|
|
|
//#include "opencv.h"
|
|
//#cgo linux pkg-config: opencv
|
|
//#cgo darwin pkg-config: opencv
|
|
//#cgo freebsd pkg-config: opencv
|
|
//#cgo windows LDFLAGS: -lopencv_core242.dll -lopencv_imgproc242.dll -lopencv_photo242.dll -lopencv_highgui242.dll -lstdc++
|
|
import "C"
|
|
import "unsafe"
|
|
|
|
const (
|
|
/* Select random initial centers in each attempt. */
|
|
KMEANS_RANDOM_CENTERS = 0
|
|
|
|
/* Use kmeans++ center initialization by Arthur and Vassilvitskii [Arthur2007]. */
|
|
KMEANS_PP_CENTERS = 2
|
|
)
|
|
|
|
/*
|
|
KMeans finds centers of k clusters in data and groups input samples around
|
|
the clusters. It returns a matrix that stores the cluster indices for every
|
|
sample, and a matrix that stores the cluster centers.
|
|
*/
|
|
func KMeans(data *Mat, k int, termcrit TermCriteria, attempts int, rng RNG, flags int) (labels, centers *Mat) {
|
|
var compactness C.double
|
|
|
|
labels = CreateMat(data.Rows(), 1, CV_32S)
|
|
centers = CreateMat(k, 1, data.Type())
|
|
|
|
C.cvKMeans2(
|
|
unsafe.Pointer(data),
|
|
C.int(k),
|
|
unsafe.Pointer(labels),
|
|
(C.CvTermCriteria)(termcrit),
|
|
C.int(attempts),
|
|
(*C.CvRNG)(&rng),
|
|
C.int(flags),
|
|
unsafe.Pointer(centers),
|
|
&compactness)
|
|
|
|
return labels, centers
|
|
}
|