Since the Semantics API is not publicly available yet, we could provide a fallback tree watering possibly based on Google Vision API annotations: * https://cloud.google.com/vision/docs/object-localizer * https://cloud.google.com/vision/docs/reference/rest/v1p2beta1/images/annotate#block