I’m currently doing a machine learning course and spending…slightly more time than usual contemplating high-dimensional data.
It’s conventional to think of images as having two dimensions. However, at least for the purposes of machine learning, they have two spatial dimensions and any number of others (including, most obviously, color).
Color may itself have several dimensions in order to account for different color channels (e.g. red, green, and blue). Or, color may contribute just one additional dimension, in the case of greyscale images.
For fun, I made a site that helps to visualize the color dimension by tacking on a third spatial dimension to the image. Drag the images the see the flip-side.