hiamitabha on Reddit shows how Vector can use the Google Vision API to recognize objects. So far the results have been mixed, but I am sure this can be improved – and it shows how to access thrid party APIs via the Python SDK.
hiamitabha provides an SDK example via Github for you to scrutinize and also a video, so head over to Reddit to read about his findings.