def clustering(NC, ref_imgs, img_infos): clusters, centroids = K.KMeans_clustering(ref_imgs, NC, img_infos) np.save(clusters, 'clusters' + NC + '.npy') np.save(centroids, 'centroids' + NC + '.npy')
j = test_files[i] imgs.append(ref_imgs[j]) input_infos.append(img_infos[j]) ref_imgs = np.delete(ref_imgs, (j), axis=0) img_infos = np.delete(img_infos, (j), axis=0) distance_matrix = np.delete(distance_matrix, (j), axis=0) distance_matrix = np.delete(distance_matrix, (j), axis=1) # loading clusters and centroids clusters2, centroids2 = np.load('clusters2.npy').item(), np.genfromtxt( 'centroids2.csv', delimiter=',') centroids2 = centroids2.astype(int) clusters3, centroids3 = np.load('clusters3.npy').item(), np.genfromtxt( 'centroids3.csv', delimiter=',') centroids3 = centroids3.astype(int) clusters11, centroids11 = K.KMeans_clustering(ref_imgs, 1, img_infos) for max_iter in range( 1, 200, 1 ): # We test the accuracy of each algorithm having different values for maximum iterations print(max_iter) #Testing the ML-ANN algorithm NC = 1 # number of clusters weights = np.zeros(NC) # weights for each cluster CDistances = np.zeros(NC) # dictances between clusters and input image distances = {} for i in range(0, NC): distances[str(i)] = [] clusters, centroids = copy.deepcopy(clusters11), copy.deepcopy( centroids11) # clustering reference images by K-Means algorithm