How to use trainNetwork in CNN for both image and feature input?

I have image set of 20X40 pixel. and it has 3 different feature and a total of 1700000 sample in each catagory.

So, overall data : 20 X 40 X 3 X 1700000

I have 5 features for each image. so total data : 5 X 1700000.

It is a regression problem and I want to use both imageInputLayer and featureInputLayer. How do I use trainNetwork in CNN for that?

Any example to train the network?