stratify=y)
X = X_test
y = y_test
else:
print("\nThe provided dataset does not exist!")
sys.exit()
if clustering_algo == 'k_means':
#time-start
start = timeit.default_timer()
#object initialization
km = my_kmeans.my_kmeans(X, y)
if n_cluster == None:
n_cluster = 3
#fit
km.fit(X, n_cluster)
#predict
predicted_y = km.predict(X)
#elbow method
km.elbow_method(X)
#time-stop
stop = timeit.default_timer()
# display the vertices & normals
ax = plt.figure().gca(projection='3d')
ax.quiver(points[:, 0], points[:, 1], points[:, 2], normals[:, 0], normals[:, 1], normals[:, 2], length=0.01)
ax.set_title("Data")
# select the number of clusters
k = 5
print("Starting computation...")
# TODO:
# Use your k-means implementation to find 'k' cluster centers in
# the normal vectors. The array 'map_to_centers' should contain the index of the
# center of each normal in 'VN'.
#
centers = np.random.standard_normal((k, 3))
centers=centers/np.linalg.norm(centers)
map_to_centers, center_colors = my_kmeans(points, k, centers)
# Cluster Visualization
colors=["r", "g", "b", "c", "m", "y", "k"]
ax = plt.figure().gca(projection='3d')
ax.set_title("Clustering")
# TODO:
# create a plot as above, where each cluster has a different color
ax.quiver(points[:, 0], points[:, 1], points[:, 2], normals[:, 0], normals[:, 1], normals[:, 2], length=0.01, colors=colors)
ax.set_title("Data")
print("done")
plt.show()
for i1 in range(len(ks)):
# number of clusters
k = ks[i1]
print "Reiterate"
# display image
plot_image(i1, 0, colors)
# plt.show()
# <exercise>
# Use your *own* k-means implementation to find 'k' cluster centers in
# the colors 'colors'. Assign the variable 'center_colors' with the
# representative color of each center (dimension: k x 3). The array
# 'map_to_centers' should contain the center of each color in 'colors'.
#
# </exercise>
map_to_centers, center_colors = my_kmeans(colors, k)
print center_colors
print map_to_centers
plot_image(i1, 1, map_to_centers)
print "Plotted"
plot_image(i1, 2, center_colors[map_to_centers])
plt.show()
data = data.reshape(img.shape[:2] + (-1,))
if data.shape[2] == 1:
data = data.reshape(img.shape[:2])
axes[i1, i2].imshow(data.astype('uint8'))
axes[i1, i2].set(xticks=[], yticks=[])
print("Starting computation...")
for i in range(len(clusterAmounts)):
# TODO:
# Use your k-means implementation to find 'k' cluster centers in
# the colors 'colors'. Assign the variable 'center_colors' with the
# representative color of each center (dimension: k x 3). The array
# 'map_to_centers' should contain the center of each color in 'colors'.
centers = np.random.random((clusterAmounts[i], 3)) * 255
map_to_centers, center_colors = my_kmeans(colors,clusterAmounts[i], centers)
# display image
plot_image(i, 0, colors) # original image
axes[i, 0].set_title("Original")
plot_image(i, 1, map_to_centers) # clusters
axes[i, 1].set_title("Clusters k={}".format(clusterAmounts[i]))
# TODO:
cluster_colors = np.empty_like(colors)
# Plot a reconstruction of the image, using only the cluster centers as colors
for k in range(clusterAmounts[i]):
cluster_colors[map_to_centers == k] = center_colors[k]
# load data: # points: (Nx3) vector of vertex coordinates # normals: (Nx3) vector of vertex normals data = scipy.io.loadmat(os.path.join(*'.. data.mat'.split())) points, normals = data['points'], data['normals'] # display the vertices & normals p_, n_ = points, normals mlab.figure() mlab.quiver3d(p_[:, 0], p_[:, 1], p_[:, 2], n_[:, 0], n_[:, 1], n_[:, 2]) # select the number of clusters k = 5 # <exercise> # Use your *own* k-means implementation to find 'k' cluster centers in # the norms. The array 'map_to_centers' should contain the index of the # center of each normal in 'VN'. # # </exercise> map_to_centers, _= my_kmeans(normals, k) # display the clustering mlab.figure() n__ = normals * (map_to_centers + 1)[:, None] mlab.quiver3d(p_[:, 0], p_[:, 1], p_[:, 2], n__[:, 0], n__[:, 1], n__[:, 2], scale_mode='none') mlab.show()
