Is it possible to use imagedatastore with jpg and vis images together and feed them separately to CNN

I have a dataset that contains RGB and Depth images with 10 classes. At first, I used the RGB images with imagedatastore and build my deep learning system. Now, I want to use both depth and RGB images by feed each into four layers route. (four for depth and four for RGB)

The thing that I confused with is, how to use imagedatastore with both of them in the same time?

Should I mix them in one related class/folder, or separate them into different folders with same label name, and then use two imagedatastore (imds1 and imds2)?

If I mixed them, how can I recognise them (depth, RGB) as they need to be fed to the layers routes.

If I separated them or used two imgedatastore functions, how can I use the training and testing as there will be two TrainingSet and two TestingSet

Best Regards