Listing 10.6 and 10.16 confusing indexing

[jxu]

As shown in Chapter 8, Keras indexes images with coordinates (height, width, depth).

In 10.6, isn't it n_cols = images_per_row, assuming n_cols means number of cols and not length of cols?

In 10.16, why is stitched_filters init as shape (full_width, full_height, 3)? The issue is compounded by the use of n for both dimensions.