Amazon Rekognition Update - More Accurate Object and Scene Detection: aws.amazon.com/about-aws/what… . Now returns bounding boxes for each object it identifies: 
You can use that info to know how many occurrences of a detected object are in an image (Here are Carlitos and Luna taking a break after a hard day of doing nothing):
And you can look at overlap and relative position of bounding box to infer relationships, such as the fact that it is me (circa 1964) wearing the suit:
The data returned for objects is now hierarchical - burgers and fries are food....
Newest docs are at docs.aws.amazon.com/rekognition/la…
