cloud_vision_nl.py

# Call the vision API to extract text from the image
imageresponse = vision_client.document_text_detection(image=image)

# Lump all the text from the image in a single string
text = ' - '.join(imageresponse.text_annotations[0].description.split('\n'))

# Call the natural language API to extract entities for our text
document = language.types.Document(content=text, type=language.enums.Document.Type.PLAIN_TEXT)
languageresponse = language_client.analyze_entities(document=document)

# Collect all text that was identified as a "PERSON"
entities = languageresponse.entities
persons = [e.name for e in entities if e.type == 1]

# Loop over all persons and draw black boxes over them
for person in persons:
    for annotation in imageresponse.text_annotations[1:]:
        if annotation.description in person:
            v = annotation.bounding_poly.vertices
                box = [v[0].x, v[0].y, v[2].x, v[2].y]
                draw.rectangle(box, fill='black')