Initial vision API addition.

[daspecster]

Fairly rough and a lot of conversation to be had.

• Face
• Landmark
• Logo
• Label
• Text
• Safe Search

Examples to play with

>>> from gcloud.vision import Client
>>> vc = Client('ferrous-arena')
>>> f = open('/tmp/car.jpg', 'r')
>>> vc.annotate(f.read(), 'LABEL_DETECTION', 3)
>>> vc.annotate('gs://ferrous-arena-my-test-bucket/car.jpg', 'LABEL_DETECTION', 3)

[theacodes]

vc.annotate(f.read(), 'LABEL_DETECTION', 3)

I'd prefer constants over strings, even if they semantically mean the same thing:

vc.annotate(f.read(), vision.LABEL_DETECTION, 3)

[dhermes]

Rather than writing and dumping all this code, please draft a vision-usage.rst doc and we can discuss features before there is code.

Once we agree on something, try to land them feature by feature so review doesn't become unmanageable. (e.g. you couldn't just sent a tiny PR with the Connection, then a tiny PR with the Client and then build up the client for each feature).

(PS I know I "broke the rules" in #1772 however the entire API is contained in three methods so I did it "on purpose" / was aware it was a bit of a brain dump.)

[tseaver]

(PS I know I "broke the rules" in #1772 however the entire API is contained in three methods so I did it "on purpose" / was aware it was a bit of a brain dump.)

#1772 has have 70+ comments on it, which is a sign that we should've argued out the API surface first. :)

[dhermes]

@tseaver I totally agree and wish I had done that.

[daspecster]

Ok, I'll make a PR with just a vision-usage.rst and then we can talk about it there.

[tseaver]

In favor of #1809 ans subsequent.