Convention when feeding data to an annotator

Hi, I have some trouble using face annotators, and I think the issue is caused by some inconsistency between feeding a batch of images vs a single image when calling the annotator. More precisely :

• The FaceCrop sends a single image to the annotator, and expects to receive a single dictionary of annotations
• While (from my understanding) the Base annotator expects a list of images and returns a list of annotations dictionaries (one dictionary per provided image).

I should be able to propose an easy fix, but which is the correct expected behavior ?

ping @ydayer @tiago.pereira