Instruments & Data Tools: solutions for sensing, imaging, analytics

Instruments & Data Tools is specialising in custom algorithms and instrumentation for spectroscopy and imaging.
We develop solutions for science and industry.

Animated 3D graphs with Matplotlib mplot3d toolkit

The The mplot3d toolkit can be used as an add-on to matplotlib for simple 3D plots and charts. In addition, the interactive backends enable rotating and zooming the 3D graphs.

Here we show how to make a very simple animation of a 3D scatter plot using the mplot3d toolkit.

In this example we are going to use data derived by Principal Components Analysis (PCA) applied to near-infrared (NIR) spectra of different samples. NIR spectra contains hundreds of data points, which can be reduced to just a handful by PCA. For more information about PCA you can check out our post here, or you can see how we used PCA to classify macadamia kernels based on their NIR spectra.

Today we are going to start with a set of data point stored in a numpy array X which contains the coordinates of the first 3 principal components worked out by PCA. For instance the first 5 points of the array X would look like this:

array([[-2.387369  , -5.37056941,  8.78933234], 
       [-3.27891755, -5.67361897,  3.61089993], 
       [ 0.91674654, -5.16642442,  2.64881256], 
       [ 3.38976338, -5.04415541,  6.92688102], 
       [-1.7956064 , -7.01851776,  8.10397704]]) 

Associated with X we have an array of labels, simply numbers referring to some classification of the sample measured. We will use the labels to color code the points of our scatter plot.

OK, without further ado let’s write the imports we need for the graph:

import matplotlib.pyplot as plt
from mpl_toolkits.mplot3d import Axes3D

And here’s how we make the plots. The idea is to generate and save plots by changing the view angle every time.

# Extract unique labels 
unique = list(set(labels))
# Define the colors
colors = [plt.cm.jet(float(i+1)/(max(unique)+1)) for i in unique]

# Define plot style
with plt.style.context(('ggplot')):

    # Loop over the view angles
    for angle in range(0,360,5):    
        
        # Define figure and 3D axes
        fig = plt.figure(figsize=(10,7))
        ax = Axes3D(fig)

        # Loop over every element of X
        for i, u in enumerate(unique):
            xi = [X[j,0] for j  in range(len(X[:,0])) if lab[j] == u]
            yi = [X[j,1] for j  in range(len(X[:,1])) if lab[j] == u]
            zi = [X[j,2] for j  in range(len(X[:,2])) if lab[j] == u]
            
            # Scatter plot
            ax.scatter(xi, yi, zi, c=colors[i], s=80, label=str(u))

        # Set the view angle
        ax.view_init(30, angle)

        # Label the axes
        ax.set_xlabel('PC1')
        ax.set_ylabel('PC2')
        ax.set_zlabel('PC3')

        # Save the figure
        plt.savefig('.\\movie\\t4_'+str(angle).zfill(3)+'.png', dpi=100)

Here’s how one of the frames looks like.

And here’s an animation generated in this way…

That’s it for today. Feel free to send us comments or suggestions using the box below. Thanks for stopping by!