In the lecture, I thought the image array had to be flattened such that all the reds, greens and blue values would be listed together, e.g. [r1, r2, r3, g1, g2, g3, b1, b2, b3].
However in the implementation (as well as the passed test using two colors), the flattened array appears to be [r1, g1, r2, g2 …]
I think your observation is correct.
Let’s create 3x2x3 ndarray like this.
As you see, the result of flatten (which should be same as our exercise that use reshape) is
array([ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17])
numpy flatten has an option to change the order. Default is “C”. If we use “F”, which is Fortran format, the result is as follows.
array([ 0, 6, 12, 3, 9, 15, 1, 7, 13, 4, 10, 16, 2, 8, 14, 5, 11, 17])
It starts with the first channel (Red in the figure), but axis is not exactly same as the image provided.
