*This is not an official GCP repository

Google Cloud ML Walkthrough Demo

Introduction

This web application integrates Google Cloud Vision, Google Cloud Speech, Google Cloud Translate, and Google Cloud Natural Language with WebRTC as a audio/video source.

Architecture

Pre-Requisites

1. Need a Google Cloud Platform account

Setup

1. Create GCP Project
2. Activate Google Cloud Shell
3. Set your session to the current project

   gcloud config set project <PROJECT NAME>
   

4. Enable Compute Engine, Vision, Speech, Translate, and Natural Language APIs

    gcloud services enable compute.googleapis.com vision.googleapis.com speech.googleapis.com translate.googleapis.com language.googleapis.com vision.googleapis.com
   

5. Create an API key

Speech Client Compute Engine

This application uses Google Cloud Speech streaming API. In order to stream audio from the web application to the Speech API, you will need to create a GCE instance to setup a secure web socket. This was originally setup using the Speaking with a Webpage codelab.

1. Create a micro GCE instance

   gcloud beta compute --project=$(gcloud config get-value project) instances create speech-client --zone=us-east1-b --machine-type=n1-standard-1 --subnet=default --scopes=https://www.googleapis.com/auth/cloud-platform --tags=https-server --image=debian-9-stretch-v20180611 --image-project=debian-cloud --boot-disk-size=10GB --boot-disk-type=pd-standard --boot-disk-device-name=speech-client
   

2. SSH into the new instance
3. Install git and maven

   sudo apt-get update
   sudo apt-get install -y maven openjdk-8-jdk git
   

4. In order to access the speech client, you will need to open a port in the Firewall

   gcloud compute firewall-rules create dev-ports --allow=tcp:443 --source-ranges=0.0.0.0/0
   

5. Clone this project

   git clone https://github.com/tfrantzen/gcp-ml-walkthrough.git
   

6. Open directory

   cd gcp-ml-walkthrough/streaming-client/
   

7. Deploy application

   sudo mvn clean jetty:run
   

ML Walkthrough Application

1. Open the Google Cloud Shell
2. Clone project

   git clone https://github.com/tfrantzen/gcp-ml-walkthrough.git
   

3. Open directory

   cd gcp-ml-walkthrough/client/
   

4. Install dependencies

   npm install
   

5. Collect GCE IP Address and API Key
6. Edit public\js\app.js, update variables on line 1 and 2

   ...
   var apiKey = '<API_KEY>';
   var speechApiIP = '<GCE_INSTANCE_EXTERNAL_IP>';
   
   
   var translateUri = `https://translation.googleapis.com/language/translate/v2?ke...
   

7. Deploy application

   npm start
   

8. Open the web preview
   • You will need to navigate your speech-client GCE instance first to accept the self signed certificate

Sources

1. https://codelabs.developers.google.com/codelabs/speaking-with-a-webpage/index.html?index=..%2F..%2Findex#0
2. https://webrtc.github.io/samples/
3. https://github.com/GoogleCloudPlatform