What does it mean for an image to have 512 channels?

From the the programming assignment of C2_W3 on StyleGan: “imagine an image has 512 channels and its height and width are (4 x 4)”. This also occured a few times in the lectures and when I inspected the paper.

My assumption is a image have a maximum of 3 (RGB) channels, and grayscale image have 1 channel, so am quite confused what having 512 means. Is it refering to 512 feature dimension? Or maybe an image with more than 3 (rgb) color dimensions?

Hi Hon,

You are right, it’s just referring to image with 512 feature dimensions and not more than 3 (rgb) color dimensions. It’s just assuming that the image has passed through a convolution layer and has the specified number of features.

This link explains in details how different filters in the convolution layer can lead to different output dimensions.

Hope this helps.

1 Like