kmeans_plot_img_compression.py

# Read the image
img = imread('images/my_image.jpg')
img_size = img.shape

# Reshape it to be 2-dimension
X = img.reshape(img_size[0] * img_size[1], img_size[2])

# Run the Kmeans algorithm
km = KMeans(n_clusters=30)
km.fit(X)

# Use the centroids to compress the image
X_compressed = km.cluster_centers_[km.labels_]
X_compressed = np.clip(X_compressed.astype('uint8'), 0, 255)

# Reshape X_recovered to have the same dimension as the original image 128 * 128 * 3
X_compressed = X_compressed.reshape(img_size[0], img_size[1], img_size[2])

# Plot the original and the compressed image next to each other
fig, ax = plt.subplots(1, 2, figsize = (12, 8))
ax[0].imshow(img)
ax[0].set_title('Original Image')
ax[1].imshow(X_compressed)
ax[1].set_title('Compressed Image with 30 colors')
for ax in fig.axes:
    ax.axis('off')
plt.tight_layout();

where did that cluster_centers_ came from in line 13.

X_compressed = km.cluster_centers_[km.labels_]

It is an attribute for "km" object. You can look at Sklearn's doc for more details.