ocr

Google Cloud Functions Recipes

OCR with Cloud Vision API and Google Translate API

Overview

This recipe shows you how to use the Cloud Vision API together with the Google Translate API using Cloud Pub/Sub as a message bus. Where applicable:

Replace [PROJECT-ID] with your Cloud Platform project ID

What's going on here?

1. Image is uploaded to Cloud Storage with text in any language (text in the image itself)
2. Cloud Function is triggered, uses the Vision API to extract the text, and the Translate API to detect the language
3. For all languages we're translating into (except the language of the text), publish a message to ther translate topic
4. For the language that matches the language of the text, bypass translation and publish to the save topic
5. Cloud Function is triggered and uses the Translate API to translate the message into various languages, then publishes each translation to the save topic
6. Cloud Function is triggered and saves text to Cloud Storage
7. Translated text from the original source image is downloaded

Cooking the Recipe

1. Follow the Cloud Functions quickstart guide to setup Cloud Functions for your project

2. Enable the Vision API and the Translate API

3. Clone this repository

   cd ~/
   git clone https://github.com/jasonpolites/gcf-recipes.git
   cd gcf-recipes
   

4. Create a Google Translate API Key

5. Create a file called translate_apikey.json and copy the Translate API Key into this file as a String

   echo "\"[YOUR API KEY]\"" > ocr/app/translate_apikey.json
   

6. Run the setup for the ocr sample:

   npm install
   node setup install ocr [PROJECT-ID]
   

7. Upload a sample image

   gsutil cp ocr/samples/sample_ch.jpg gs://[PROJECT-ID]-gcf-samples-ocr-in/ 
   

8. Watch the logs to make sure the executions have completed

   gcloud alpha functions get-logs --limit 100
   

9. Pull the extracted text from the bucket and pipe to standard out

   gsutil cat gs://[PROJECT-ID]-gcf-samples-ocr-out/sample_ch_to_en.txt
   

Running Tests

This recipe comes with a suite of unit tests. To run the tests locally, just use npm test

npm install
npm test

The tests will also produce code coverage reports, written to the /coverage directory. After running the tests, you can view coverage with

open coverage/lcov-report/index.html