Misleading axis interpretation in SVD tutorial

[StefanieSenger]

The SVD tutorial interprets the shape of the image in the following way:

img.shape

(768, 1024, 3)

The output is a tuple with three elements, which means that this is a three-dimensional array. In fact, since this is a color image, and we have used the imread function to read it, the data is organized in three 2D arrays, representing color channels (in this case, red, green and blue - RGB). You can see this by looking at the shape above: it indicates that we have an array of 3 matrices, each having shape 768x1024.

I think it is misleading to explain the organisational structure of the array as consisting of three 2D arrays. The data is not organized as an array of 3 matrices.

In numpy, the leftmost axis is the outermost. In shape (768, 1024, 3) the colour channels axis stands for the innermost axis.
So rather than "three matrices of shape 768x1024," I think it helps building intuition better to talk about a 768x1024 grid of pixels, where each pixel has 3 values for RBG.

Do others agree with this? If so, I could make a PR that suggests a better wording.

[melissawm]

I think that's a good explanation - if you send in a PR I'm happy to review and merge. Thanks!