best_u = np.loadtxt(filename_template.format(tag='u'), delimiter=",")
best_u_ew = np.loadtxt(filename_template_ew.format(tag='u'), delimiter=",")
cluster_ew = np.argmax(best_u_ew, axis=1)
cluster = np.argmax(best_u, axis=1)
N, ND = data.shape
knn = 8
#create neighbourdhood EW
print("building neighbourhood, location.shape=", locations.shape)
kdtree = cKDTree(locations_ore)
neighbourhood, distances = make_neighbourhood(kdtree,
locations_ore,
knn,
max_distance=np.inf)
distances = np.array(distances)
#spatial EW
verbose_level = 2
clusters_graph_ew = graph_cut(locations_ore,
neighbourhood,
best_u_ew,
unary_constant=100.0,
smooth_constant=80.0,
verbose=1)
for i in range(N):
print(i,
cluster_ew[i],
target = False
force = False
file_template = '../results/bm_{set}_swfc_%d.csv'%NC
best_centroids = np.loadtxt(file_template.format(set='centroids'),delimiter=",")
best_weights = np.loadtxt(file_template.format(set='weights'),delimiter=",")
best_u = np.loadtxt(file_template.format(set='u'),delimiter=",")
clusters = np.argmax(best_u,axis=1)
N,ND = data.shape
knn = 15
#create neighbourdhood
print("building neighbourhood, location.shape=",locations.shape)
kdtree = cKDTree(locations)
neighbourhood,distances = make_neighbourhood(kdtree,locations,knn,max_distance=2.0)
distances = np.array(distances)
#spatial
verbose_level = 2
clusters_graph = np.int32(graph_cut(locations,neighbourhood,best_u,unary_constant=70.0,smooth_constant=15.0,verbose=1))
#for i in range(N):
# print(i,clusters[i],best_u[i],clusters_graph[i],sep=',')
np.savetxt("../results/final_bm_clusters_swfc_%d.csv"%NC,clusters_graph,delimiter=",",fmt="%d")
#filter just 0-50
#indices = np.where(locations[:,0] <= 50)[0]
#indices = np.where(locations[indices,1] <= 50)[0]
#locations = locations[indices,:]
#X = X[indices,:]
N, ND = X.shape
NC = 4
knn = 8
#create neighbourdhood
print("building neighbourhood")
kdtree = cKDTree(locations)
neighbourhood, distances = make_neighbourhood(kdtree, locations, knn)
distances = np.array(distances)
#create U fake
#best_u = np.zeros((N,NC))
#
#for i in range(N):
# best_u[i,int(X[i,6])] = 1.0
#spatial
verbose_level = 2
clusters_graph = graph_cut(locations,
neighbourhood,
best_u,
unary_constant=100.0,
smooth_constant=100.0,