def score(image_name, K):
image = plt.imread(image_name)
rows = image.shape[0]
cols = image.shape[1]
pixels = np.zeros((rows * cols, 3))
for i in range(rows):
for j in range(cols):
pixels[j * rows + i, :] = image[i, j, :]
class1, centroid1 = my_kmeans(pixels, K)
class2, centroid2 = my_kmedoids(pixels, K)
converted_image1 = np.zeros((rows, cols, 3))
converted_image2 = np.zeros((rows, cols, 3))
for i in range(rows):
for j in range(cols):
converted_image1[i, j, :] = centroid1[class1[j * rows + i], :]
converted_image2[i, j, :] = centroid2[class2[j * rows + i], :]
converted_image1 /= 255
converted_image2 /= 255
f = plt.figure()
f.set_figheight(30)
f.set_figwidth(30)
plt.subplot(1, 3, 1)
plt.title('Origin')
plt.imshow(image)
plt.axis('off')
plt.subplot(1, 3, 2)
plt.title('K-means')
plt.imshow(converted_image1)
plt.axis('off')
plt.subplot(1, 3, 3)
plt.title('K-medoids')
plt.imshow(converted_image2)
plt.axis('off')
plt.show()
return None
def my_kmedoids(image_data, K, threshold=0):
N = image_data.shape[0]
p = image_data.shape[1]
medoids = np.zeros((K, p))
_, medoids = my_kmeans(image_data, K)
medoids = np.asarray(medoids)
medoids_old = np.zeros(medoids.shape)
medoids_new = deepcopy(medoids)
error = np.linalg.norm(medoids_new - medoids_old)
image_data = np.asarray(image_data)
labels = np.zeros(N)
DisMat = np.zeros((N, K))
iter_ct = 0
while error > threshold:
iter_ct += 1
# print('K-medoids iteration {}'.format(iter_ct))
medoids_old = deepcopy(medoids_new)
for i in range(K): # assign image_data points to closest centroids
DisMat[:, i] = np.linalg.norm(image_data - medoids_new[i],
axis=1,
ord=2)
labels = np.argmin(DisMat, axis=1)
for i in range(K):
cluster = image_data[labels == i]
DMC = sum(np.linalg.norm(cluster - medoids_new[i], axis=1, ord=2))
DMP = np.zeros(cluster.shape[0])
if cluster.shape[0] == 0:
medoids_new[i] = medoids_old[i]
else:
for j in range(cluster.shape[0]):
DMP[j] = np.sum(
np.linalg.norm(cluster - cluster[j], axis=1, ord=2))
small_cost_idx = np.argmin(DMP)
if DMP[small_cost_idx] < DMC:
medoids_new[i] = cluster[small_cost_idx]
error = np.linalg.norm(medoids_new - medoids_old)
return labels.astype(int), medoids