CODEWITHSUNDEEP
google-cloud-storagegoogle-cloud-vision
Nilesh Bhosale
Nilesh Bhosale

Reputation: 35

Google Vision API detects labels for only public images on GCP

I am storing images on Google Cloud Storage and using Google Vision APIs to detect labels of those images. I use the same account and credentials for both purposes. I am using the sample program given at: 'https://github.com/GoogleCloudPlatform/python-docs-samples/blob/master/vision/cloud-client/detect/detect.py'

I can successfully detect labels for the local images and images on internet which are publicly accessible. When I use the following with a image stored in a bucket on my GCP storage, the program does not detect any labels unless I mark the data (image) as public.

e.g.

When it is private:

# ./detect.py labels-uri 
'https://www.googleapis.com/download/storage/v1/b/mybucket/o/Penguins.jpg?
generation=1510548912343529&alt=media'
Labels:

When I mark it as 'public':

# ./detect.py labels-uri 
'https://www.googleapis.com/download/storage/v1/b/mybucket/o/Penguins.jpg?
generation=1510548912343529&alt=media'
Labels:
penguin
bird
king penguin
flightless bird
beak
organism

I was expecting, since I am using the same credentials for the vision and storage API access, it should even work on my private images.

Can you help?

Upvotes: 1

Views: 871

Answers (2)

Emilia Apostolova
Emilia Apostolova

Reputation: 1787

This worked for me with a non-public GS bucket:

export GOOGLE_APPLICATION_CREDENTIALS="path-to-your-creds.json"

Python3:

import json

import google.auth
import requests
from google.auth.transport import requests as grequests
from google.cloud import storage

URL = "https://vision.googleapis.com/v1p4beta1/files:asyncBatchAnnotate"


credentials, project_id = google.auth.default(scopes=["https://www.googleapis.com/auth/cloud-platform"])
credentials.refresh(grequests.Request())
headers = {'Authorization': 'Bearer ' + credentials.token, "Content-Type": "application/json; charset=utf-8"}

request_json = {
    "requests": [
        {
            "inputConfig": {
                "gcsSource": {
                    "uri": "gs://your-input-bucket/test.pdf"
                },
                "mimeType": "application/pdf"
            },
            "features": [
                {
                    "type": "DOCUMENT_TEXT_DETECTION"
                }
            ],
            "outputConfig": {
                "gcsDestination": {
                    "uri": "gs://your-output-bucket/json"
                },
                "batchSize": 1
            }
        }
    ]
}

response = requests.post(URL, json=request_json, headers=headers)
print(response.content)

Upvotes: 0

Brandon Yarbrough
Brandon Yarbrough

Reputation: 38389

When referencing a Cloud Storage object, use the URI pattern gs://bucket_name/object_name.

Try ./detect.py labels-uri gs://mybucket/Penguins.jpg

Cloud Vision supports both Cloud Storage objects as well as any arbitrary URL. However, when you reference a URL, Cloud Vision does not forward your credentials there, unlike when you reference a Cloud Storage object directly. Here you're specifying a URL that would attempt to download a Cloud Storage object anonymously, which is not what you want. (Note however that Cloud Vision does not support specifying a specific version of a GCS object -- for more, see https://cloud.google.com/vision/docs/reference/rest/v1/images/annotate).

Upvotes: 2

Related Questions