visioneventing-python.md

Create Vision Handler - Python

Create Vision Handler

Adapt the previous message dumper sample by creating a new app.py:

from flask import Flask, request
...
@app.route('/', methods=['POST'])
def storage_event():
    attr = json.loads(request.data)['Attributes']
    if attr['eventType'] == 'OBJECT_FINALIZE':
        analyze_image(attr['bucketId'], attr['objectId'])
    return 'OK', 200

Note: When the file is created/updated, an event of type OBJECT_FINALIZE is received. Bucket and file names are respectively given by bucketId and objectId attributes.

Handle Image Analysis

Once we have the file info, we can pass it to the Vision API:

from google.cloud import vision
...
def analyze_image(bucket_id, filename):
    client = vision.ImageAnnotatorClient()
    image = vision.types.Image()
    image.source.image_uri = f'gs://{bucket_id}/{filename}'
    response = client.label_detection(image=image)

    annots = response.label_annotations
    labels = ', '.join([a.description for a in annots if 0.5 <= a.score])
    info(f'Picture labels: {labels}')

Create Dockerfile

Create a Dockerfile for the image:

FROM python:3.7-slim

RUN pip install Flask gunicorn google.cloud.vision

WORKDIR /app
COPY . .

CMD exec gunicorn --bind :$PORT --workers 1 --threads 8 app:app

Note: google.cloud.vision client library is specified in addition to Flask & gunicorn.

What's Next?

Back to Integrate with Vision API