As Thomas Lins Pedersen explains in a recent blog post, the trick is in using the
persp function to translate points in 3-D space into a 2-D projection. This function is normally used to render a 3-D scatterplot or wireframe plot, but if you instead capture its output value, it returns a transformation matrix. You can then use the
trans3d function to with this matrix to transform points in 3-D space. Thomas demonstrates how you can pass the transformed 2-D coordinates to plot a 3-D cube, and even animate it from two slightly different perspectives to create a 3-D stereo pair:
Rendering 3-D images isn't just for fun: there's plenty of 3-D data to analyze and visualize, too. Giora Simchoni used R to visualize data from the Carnegie-Mellon Graphics Lab Motion Capture Database. This data repository provides the output of human figures in motion-capture suits performing actions like walking, jumping, and even dancing. Since the motion-capture suits include multiple sensors measured over a time-period, the data structures are quite complex. To make things simpler, Giora created the mocap package (available on Github) to read these motion data files and generate 3-D animations to visualize them. For example, here's the output from two people performing the Charleston together:
You can find complete details behind both animations, including the associated R code, at the links below.
Data Imaginist: I made a 3D movie with ggplot2 once – here's how I did it
Giora Simchoni: Lambada! (The mocap Package)