Google Cloud Vision - Which region does Google upload the images to?

Question

I am building an OCR based solution to extract information from certain financial documents. As per the regulation in my country (India), this data cannot leave India.

Is it possible to find the region where Google Cloud Vision servers are located? Alternately, is it possible to restrict the serving region from the GCP console?

This is what I have tried:

1. I went through GCP Data Usage FAQ: https://cloud.google.com/vision/docs/data-usage

2. GCP Terms of Service: https://cloud.google.com/terms/ (Look at point 1.4 Data Location on this page)

3. Talking to the GCP Sales rep. He did not know the answer.

I know that I can talk to Google support but that requires $100 to activate, which is a lot for for me.

Any help would be appreciated. I went through the documentation for Rekognition as well but it seems to send some data outside for training so not considering it at the moment.

PS - Edited to make things I have tried clearer.

Quintilla

How about this? (cloud.google.com/about/locations). They appear to be Mumbai, – Meraree

We do use the GCP servers based in Mumbai. But Cloud Vision is a separate API and I haven't found any documentation clarifying that it is served from the same region. – Quintilla

I know this is a revolutionary concept but... have you thought to ask Google outright? – Meraree

Yes. I did talk to their sales team. The executive I was assigned did not know the answer. He suggested I talk to support. But that requires money ($100) to begin with, which is a lot for a bootstrapped startup like mine. – Quintilla

Hmm. Possibly you spoke to the wrong executive. If someone on StackOverFlow were to state categorically that there are no Vision servers, or that there are Vision servers and they are located in India, or that the Vision servers are not located in India... regardless of the answer, what degree of certainty can you take from that. – Meraree

I would obviously hope that anyone answering would also cite some documentation or personal communication with Google. Also, since this forum is pointed out on the GCP support page, one of their team members might be kind enough to turn up and give an authoritative answer. Anyway, I am continuing to look for answers elsewhere. – Quintilla

Indeed - found this on 'cloud-vision-discuss' under Temporary Storage Location of uploaded documents, dated 6 December 2018, "Images are only stored in memory. Regionalization is not available yet, but stay tuned, we're looking into it next quarter." – Meraree

Thanks a lot. This should help. Will put this as the answer. – Quintilla

Answer 1

For anyone looking at this topic recently, Google Vision has introduced multi-region support in December 2019, as can be see in their release notes.

Currently Google Vision supports 2 different processing regions: eu and us, and they say that using a specific endpoint guarantees that processing will only take place in the chosen territory.

The documentation for regionalization mentions that you can simply replace the default API endpoint vision.googleapis.com with either of the regional ones:

• eu-vision.googleapis.com
• us-vision.googleapis.com

The vision client libraries offer options for selecting the endpoint as well, and the documentation gives code samples.

For example, here is how you would do it in Python:

from google.cloud import vision

client_options = {'api_endpoint': 'eu-vision.googleapis.com'}
client = vision.ImageAnnotatorClient(client_options=client_options)

Mellicent

Answer 2

As pointed out by @Tedinoz in a comment above, the answer can be found here: https://groups.google.com/forum/#!topic/cloud-vision-discuss/at43gnChLNY

To summarise: 1. Google stores images uploaded to Cloud Vision only in memory 2. Data is not restricted to a particular region (as of Dec 6, 2018) 3. They might add data residency features in Q1, 2019.

Quintilla