I am currently looking into how to tell the CNN from which month an image is. I want to use one of the resnet networks for image classification. I know that for my case the month is a very important feature.
I came across an interesting article on the nvidia website.
Especially the third approach looks interesting, but it results in 12 time feature points per day.
My original idea was to create an additional channel that takes the month as input (in case of a single value I wanted simply resize an array to match the image shape and have that value at every point), but I am not sure how I would do that with 12 values.
Does anyone have an idea?