Preparing training data for crop species classification from drone images

I have high resolution multispectral (R,G,B,NIR,RE bands) images of a field taken from MicaSense RedEdge mounted on a drone. There are various species of crops planted. I want to classify the crops or crop species somewhat like this.

Image ref: https://www.intechopen.com/books/processing-and-analysis-of-hyperspectral-data/hyperspectral-image-classification

I am new to this. This is a classification task, so I’ll also have to label the data. How should I label the data so that it can be fed to a classification mechanism, say CNN or SVM? What would the final X -> y mapping of training data look like?

I think we can use ArcGIS for this. Still, I’d like to know the manual labeling procedure and what is done under the hood just to gain some more understanding.

Do I have to create bounding boxes of some size around the crops, like what’s done in object detection?