[RUBY] [Rails] How to detect radical images by analyzing images using Cloud Vision API

Target

If you edit the user's profile image with refile Image analysis of the uploaded image with Cloud Vision API In the case of a radical image, it cannot be saved in the DB.

Development environment

・ Ruby: 2.5.7 Rails: 5.2.4 ・ Vagrant: 2.2.7 -VirtualBox: 6.1 ・ OS: macOS Catalina

Premise

The following has been implemented.

-Login function implementation ・ Implementation of posting function

Enable Cloud Vision API

1. Activate the Cloud Vision API by following the same procedure as "Enable Geocoding API" in the following article.

How to calculate latitude / longitude using Geolocation API

2. Go to the home screen and click ʻIAM and Manage`

3. Click Service Account

4. Click Add Service Account

5. Enter the service account name (appropriate) and click Create

6. Click Continue

7. Click Create Key

8. Set the key type to JSON and click Create

9. Confirm that the key has been created and click Close

10. Click Done

11. Move the key directly under the application

About the evaluation of Cloud Vision API

Evaluation category

Evaluation value

Implementation

1. Introduce Gem

Regardless of the evaluation category, the evaluation value is LIKELY or VERY_LIKELY If it is included, we will implement it so that it will not be saved in the DB.

Gemfile

gem 'google-cloud-vision'

Terminal

$ bundle

2. Create and edit the Cloud Vision API configuration file

Terminal

$ touch lib/vision.rb

lib/vision.rb

require 'base64'
require 'json'
require 'net/https'
module Vision
  class << self
    def image_analysis(profile_image)
      image_annotator = Google::Cloud::Vision::ImageAnnotator.new(
          version: :v1,
          credentials: JSON.parse(File.open('Key file name') do |f| f.read end)
      )

      #Request parameter creation
      image = profile_image #Image you want to analyze(argument)
      requests_content = { image: { content: image }, features: [{ type: :SAFE_SEARCH_DETECTION }] }
      requests =   [requests_content]

      #Send images to Cloud Vision API
      response = image_annotator.batch_annotate_images(requests)
      result = response.responses[0].safe_search_annotation.to_h

      #The analysis result is "LIKELY" or "VERY"_Returns false if "LIKELY" is included, true otherwise
      if result.values.include?(:LIKELY) || result.values.include?(:VERY_LIKELY)
        return false
      else
        return true
      end
    end
  end
end

3. Edit controller

users_controller.rb

def update
  #When the image is edited
  if params[:user][:profile_image].present?
    #parameter(image)As a "tempfile" and assign it to a variable
    profile_image = File.open(params[:user][:profile_image].tempfile)
    #Image analysis with Cloud Vision API and assign the analysis result to a variable
    result = Vision.image_analysis(profile_image)
  else
    #Substitute "true" if the image has not been edited
    result = true
  end
  #Conditional branching depending on the analysis result
  if result == true
    @user.update(user_params)
    redirect_to user_user_path(@user)
  elsif result == false
    flash[:notice] = 'Image is inappropriate'
    render 'edit'
  end
end