void Clusteringclass::radial_basis_kmean()
{
typedef dlib::matrix<float> sample_type;
typedef dlib::radial_basis_kernel<sample_type> kernel_type;
dlib::kcentroid< kernel_type > kc;
dlib::kkmeans< kernel_type > kmean(kc);
std::vector<sample_type> initial_centers;
// tell the kkmeans object we made that we want to run k-means with k_clusters sets
kmean.set_number_of_centers(_k_clusters);
// You need to pick some initial centers for the k-means algorithm. So here
// we will use the dlib::pick_initial_centers() function which tries to find
// n points that are far apart (basically).
pick_initial_centers(_k_clusters, initial_centers, _input_data, kmean.get_kernel());
// now run the k-means algorithm on our set of samples.
kmean.train(_input_data, initial_centers);
// now loop over all our samples and get their predicted class.
for (unsigned long i = 0; i < _input_data.size(); ++i)
{
_assigned_cluster.push_back(kmean(_input_data[i]));
std::cout << _input_data[i] << std::endl;
std::cout << _assigned_cluster[i] << std::endl;
}
}
int main(int argc, char** argv){
IplImage* img = cvLoadImage(argv[1],CV_LOAD_IMAGE_COLOR);
IplImage* img_apply = cvCreateImage(cvSize(img->width,img->height),8,3);
cvNamedWindow("cluster demo",1);
unsigned char* imgData = (unsigned char*)img->imageData;
unsigned char* img_apply_data = (unsigned char*)img_apply->imageData;
Cluster* cluster=NULL;
int cl_num = initCluster(&cluster,atoi(argv[2]));
for (int i=0;i<30;i++){
kmean(imgData,img->width*img->height,cluster, cl_num);
dumpCluster(cluster,5);
kmeanApply(img_apply_data,imgData,img->width*img->height,cluster,cl_num);
cvShowImage("cluster demo",img_apply);
cvWaitKey(100);
}
cvWaitKey(0);
return 0;
}
int shotodol_kmeans_cluster(char*infile, char*outfile, int kval, int*ecode) {
struct netpbm_image img, img_out;
netpbm_init_img(&img);
netpbm_init_img(&img_out);
int i;
img.filename = infile;
img_out.filename = outfile;
if(outfile == NULL || infile == NULL) {
*ecode = 526;
return -1;
}
if(netpbm_open_and_read(&img, ecode) == -1) {
return -1;
}
if(img.type != NETPBM_IMAGE_TYPE_PPM) {
*ecode = 512;
netpbm_destroy(&img);
return -1;
}
if(netpbm_alloc_from_src(&img_out, &img)) {
goto cleanup_memory;
}
//printf("*** performing kmeans cluster ***\n");
Cluster* cluster=NULL;
int cl_num = initCluster(&cluster, kval);
int total = img.width*img.height;
for (i=0;i<30;i++){
kmean(img.pixels.color, total, cluster, cl_num);
dumpCluster(cluster,5);
kmeanApply(img_out.pixels.color, img.pixels.color, total,cluster,cl_num);
}
netpbm_write(&img_out, NULL);
cleanup_memory:
netpbm_destroy(&img);
netpbm_destroy(&img_out);
return 0;
}
