Pic-a-daily: Lab 1—Store and analyse pictures

1. Overview

In the first code lab, you will upload pictures in a bucket. This will generate a file creation event that will be handled by a function. The function will make a call to Vision API to do image analysis and save results in a datastore.

What you'll learn

• Cloud Storage
• Cloud Functions
• Cloud Vision API
• Cloud Firestore

2. Setup and Requirements

Self-paced environment setup

1. Sign-in to the Google Cloud Console and create a new project or reuse an existing one. If you don't already have a Gmail or Google Workspace account, you must create one.

• The Project name is the display name for this project's participants. It is a character string not used by Google APIs. You can update it at any time.
• The Project ID must be unique across all Google Cloud projects and is immutable (cannot be changed after it has been set). The Cloud Console auto-generates a unique string; usually you don't care what it is. In most codelabs, you'll need to reference the Project ID (it is typically identified as PROJECT_ID). If you don't like the generated ID, you may generate another random one. Alternatively, you can try your own and see if it's available. It cannot be changed after this step and will remain for the duration of the project.
• For your information, there is a third value, a Project Number which some APIs use. Learn more about all three of these values in the documentation.

Caution: A project ID must be globally unique and cannot be used by anyone else after you've selected it. You are the only user of that ID. Even if a project is deleted, the ID can never be used again

Note: If you're using a Gmail account, you can leave the default location set to No organization. If you're using a Google Workspace account, then choose a location that makes sense for your organization.

2. Next, you'll need to enable billing in the Cloud Console to use Cloud resources/APIs. Running through this codelab shouldn't cost much, if anything at all. To shut down resources so you don't incur billing beyond this tutorial, you can delete the resources you created or delete the whole project. New users of Google Cloud are eligible for the $300 USD Free Trial program.

Start Cloud Shell

While Google Cloud can be operated remotely from your laptop, in this codelab you will be using Google Cloud Shell, a command line environment running in the Cloud.

From the Google Cloud Console, click the Cloud Shell icon on the top right toolbar:

It should only take a few moments to provision and connect to the environment. When it is finished, you should see something like this:

This virtual machine is loaded with all the development tools you'll need. It offers a persistent 5GB home directory, and runs on Google Cloud, greatly enhancing network performance and authentication. All of your work in this codelab can be done within a browser. You do not need to install anything.

3. Enable APIs

For this lab, you will be using Cloud Functions and Vision API but first they need to be enabled either in Cloud Console or with gcloud.

To enable Vision API in Cloud Console, search for Cloud Vision API in the search bar:

You will land on the Cloud Vision API page:

Click the ENABLE button.

Alternatively, you can also enable it Cloud Shell using the gcloud command line tool.

Inside Cloud Shell, run the following command:

gcloud services enable vision.googleapis.com

You should see the operation to finish successfully:

Operation "operations/acf.12dba18b-106f-4fd2-942d-fea80ecc5c1c" finished successfully.

Enable Cloud Functions as well:

gcloud services enable cloudfunctions.googleapis.com

4. Create the bucket (console)

Create a storage bucket for the pictures. You can do this from Google Cloud Platform console ( console.cloud.google.com) or with gsutil command line tool from Cloud Shell or your local development environment.

Note: In this step, you will create the bucket from the console. However, if you prefer the command line, you can skip to the next step.

Navigate to Storage

From the "hamburger" (☰) menu, navigate to the Storage page.

Name your bucket

Click on the CREATE BUCKET button.

Note: You need to create a globally unique bucket name (here: uploaded-pictures, but you will have to choose a globally unique name).

Use uploaded-pictures as a prefix, and append your project ID as a suffix. If your project ID is my-project, call the bucket uploaded-pictures-my-project — that's the convention used in the command-line approach of the next section.

Click CONTINUE.

Choose Location

Create a multi-regional bucket in the region of your choice (here Europe).

Note: We should use the same region as the region we'll be using in subsequent labs for our other resources (Cloud Functions, App Engine, Cloud Run, etc.) for the best performance and avoid any extra ingress/egress costs.

Click CONTINUE.

Choose default storage class

Choose the Standard storage class for your data.

Click CONTINUE.

Set Access Control

As you will be working with publicly accessible images, you want all our pictures stored in this bucket to have the same uniform access control.

Choose the Uniform access control option.

Click CONTINUE.

Set Protection/Encryption

Keep default (Google-managed key), as you won't use your own encryption keys.

Click CREATE, to eventually finalize our bucket creation.

Add allUsers as storage viewer

Go to the Permissions tab:

Add an allUsers member to the bucket, with a role of Storage > Storage Object Viewer, as follows:

Click SAVE.

5. Create the bucket (gsutil)

Note: If you already created the bucket from the console in the previous step, you can skip this step.

You can also use the gsutil command line tool in Cloud Shell to create buckets.

In Cloud Shell, set a variable for the unique bucket name. Cloud Shell already has GOOGLE_CLOUD_PROJECT set to your unique project id. You can append that to the bucket name.

For example:

export BUCKET_PICTURES=uploaded-pictures-${GOOGLE_CLOUD_PROJECT}

Create a standard multi-region zone in Europe:

gsutil mb -l EU gs://${BUCKET_PICTURES}

Ensure uniform bucket level access:

gsutil uniformbucketlevelaccess set on gs://${BUCKET_PICTURES}

Make the bucket public:

gsutil iam ch allUsers:objectViewer gs://${BUCKET_PICTURES}

If you go to Cloud Storage section of the console, you should have a public uploaded-pictures bucket:

Test that you can upload pictures to the bucket and the uploaded pictures are publicly available, as explained in the previous step.

6. Test public access to the bucket

Going back to the storage browser, you'll see your bucket in the list, with "Public" access (including a warning sign reminding you that anyone has access to the content of that bucket).

Your bucket is now ready to receive pictures.

If you click on the bucket name, you'll see the bucket details.

There, you can try the Upload files button, to test that you can add a picture to the bucket. A file chooser popup will ask you to select a file. Once selected, it'll be uploaded to your bucket, and you will see again the public access that has been automatically attributed to this new file.

Along the Public access label, you will also see a little link icon. When clicking on it, your browser will navigate to the public URL of that image, which will be of the form:

https://storage.googleapis.com/BUCKET_NAME/PICTURE_FILE.png

With BUCKET_NAME being the globally unique name you have chosen for your bucket, and then the file name of your picture.

By clicking on the check box along the picture name, the DELETE button will be enabled, and you can delete this first image.

7. Create the function

In this step, you create a function that reacts to picture upload events.

Visit the Cloud Functions section of the Google Cloud console. By visiting it, the Cloud Functions service will be automatically enabled.

Click on Create function.

Choose a name (eg. picture-uploaded) and the Region (remember to be consistent with the region choice for the bucket):

There are two kinds of functions:

• HTTP functions which can be invoked via a URL (ie. a web API),
• Background functions which can be triggered by some event.

You want to create a background function that is triggered when a new file is uploaded to our Cloud Storage bucket:

You are interested in the Finalize/Create event type, which is the event that is triggered when a file is created or updated in the bucket:

Select the bucket created before, to tell Cloud Functions to be notified when a file is created / updated in this particular bucket:

Click Select to choose the bucket you created earlier, and then Save

Before you click Next, you can expand and modify the defaults (256 MB memory) under Runtime, build, connections and security settings and update it to 1GB.

After clicking Next, you can tune the Runtime, Source code and entry point.

Keep the Inline editor for this function:

Select one of the Node.js runtimes:

The source code consists of an index.js JavaScript file, and a package.json file that provides various metadata and dependencies.

Leave the default snippet of code: it logs the file name of the uploaded picture:

For now, keep the name of the function to execute to helloGCS, for testing purposes.

Click on Deploy to create and deploy the function. Once the deployment succeeded, you should see a green-circled check mark in the list of functions:

8. Test the function

In this step, test that the function responds to storage events.

From the "hamburger" (☰) menu, navigate back to the Storage page.

Click on the images bucket, and then on Upload files to upload an image.

Navigate again within the cloud console to go to the Logging > Logs Explorer page.

In the Log Fields selector, select Cloud Function to see the logs dedicated to your functions. Scroll down through the Log Fields and you can even select a specific function to have a finer-grained view of the functions related logs. Select the picture-uploaded function.

You should see the log items mentioning the creation of the function, the start and end times of the function, and our actual log statement:

Our log statement reads: Processing file: pic-a-daily-architecture-events.png, meaning that the event related to the creation and storage of this picture has indeed been triggered as expected.

9. Prepare the database

You will store information about the picture given by the Vision API into the Cloud Firestore database, a fast, fully managed, serverless, cloud-native NoSQL document database. Prepare your database by going to the Firestore section of the Cloud Console:

Two options are offered: Native mode or Datastore mode. Use the native mode, which offers extra features like offline support and real-time synchronization.

Click on SELECT NATIVE MODE.

Pick a multi-region (here in Europe, but ideally at least the same region your function and storage bucket are).

Click the CREATE DATABASE button.

Once the database is created, you should see the following:

Create a new collection by clicking the + START COLLECTION button.

Name collection pictures.

You don't need to create a document. You'll add them programmatically as new pictures are stored in Cloud Storage and analysed by the Vision API.

Click Save.

Firestore creates a first default document in the newly created collection, you can safely delete that document as it doesn't contain any useful information:

The documents that will be created programmatically in our collection will contain 4 fields:

• name (string): the file name of the uploaded picture, which is also he key of the document
• labels (array of strings): the labels of recognised items by the Vision API
• color (string): the hexadecimal color code of the dominant color (ie. #ab12ef)
• created (date): the timestamp of when this image's metadata was stored
• thumbnail (boolean): an optional field that will be present and be true if a thumbnail image has been generated for this picture

As we will be searching in Firestore to find pictures that have thumbnails available, and sorting along the creation date, we'll need to create a search index.

You can create the index with the following command in Cloud Shell:

gcloud firestore indexes composite create \
  --collection-group=pictures \
  --field-config field-path=thumbnail,order=descending \
  --field-config field-path=created,order=descending

Or you can also do it from the Cloud Console, by clicking on Indexes, in the navigation column on the left, and then creating a composite index as shown below:

Click Create. Index creation can take a few minutes.

10. Update the function

Move back to the Functions page, to update the function to invoke the Vision API to analyze our pictures, and to store the metadata in Firestore.

From the "hamburger" (☰) menu, navigate to the Cloud Functions section, click on the function name, select the Source tab, and then click the EDIT button.

First, edit the package.json file which lists the dependencies of our Node.JS function. Update the code to add the Cloud Vision API NPM dependency:

{
  "name": "picture-analysis-function",
  "version": "0.0.1",
  "dependencies": {
    "@google-cloud/storage": "^1.6.0",
    "@google-cloud/vision": "^1.8.0",
    "@google-cloud/firestore": "^3.4.1"
  }
}

Note: You'll find information about the NPM module in the Vision NPM module page, and also details about the API in the documentation. For Firestore, you'll find the NPM module page, and more information about the API in the documentation and API reference.

Now that the dependencies are up-to-date, you are going to work on the code of our function, by updating the index.js file.

Replace the code in index.js with the code below. It will be explained in the next step.

Note: The code is hosted on GitHub as well, if you want to copy/paste from there instead

const vision = require('@google-cloud/vision');
const Storage = require('@google-cloud/storage');
const Firestore = require('@google-cloud/firestore');

const client = new vision.ImageAnnotatorClient();

exports.vision_analysis = async (event, context) => {
    console.log(`Event: ${JSON.stringify(event)}`);

    const filename = event.name;
    const filebucket = event.bucket;

    console.log(`New picture uploaded ${filename} in ${filebucket}`);

    const request = {
        image: { source: { imageUri: `gs://${filebucket}/${filename}` } },
        features: [
            { type: 'LABEL_DETECTION' },
            { type: 'IMAGE_PROPERTIES' },
            { type: 'SAFE_SEARCH_DETECTION' }
        ]
    };

    // invoking the Vision API
    const [response] = await client.annotateImage(request);
    console.log(`Raw vision output for: ${filename}: ${JSON.stringify(response)}`);

    if (response.error === null) {
        // listing the labels found in the picture
        const labels = response.labelAnnotations
            .sort((ann1, ann2) => ann2.score - ann1.score)
            .map(ann => ann.description)
        console.log(`Labels: ${labels.join(', ')}`);

        // retrieving the dominant color of the picture
        const color = response.imagePropertiesAnnotation.dominantColors.colors
            .sort((c1, c2) => c2.score - c1.score)[0].color;
        const colorHex = decColorToHex(color.red, color.green, color.blue);
        console.log(`Colors: ${colorHex}`);

        // determining if the picture is safe to show
        const safeSearch = response.safeSearchAnnotation;
        const isSafe = ["adult", "spoof", "medical", "violence", "racy"].every(k => 
            !['LIKELY', 'VERY_LIKELY'].includes(safeSearch[k]));
        console.log(`Safe? ${isSafe}`);

        // if the picture is safe to display, store it in Firestore
        if (isSafe) {
            const pictureStore = new Firestore().collection('pictures');
            
            const doc = pictureStore.doc(filename);
            await doc.set({
                labels: labels,
                color: colorHex,
                created: Firestore.Timestamp.now()
            }, {merge: true});

            console.log("Stored metadata in Firestore");
        }
    } else {
        throw new Error(`Vision API error: code ${response.error.code}, message: "${response.error.message}"`);
    }
};

function decColorToHex(r, g, b) {
    return '#' + Number(r).toString(16).padStart(2, '0') + 
                 Number(g).toString(16).padStart(2, '0') + 
                 Number(b).toString(16).padStart(2, '0');
}

Note: Note that the main function name changed to vision_analysis. In deploying the function later, you will need to make sure the Entry point text field below the code is set to vision_analysis so the correct JavaScript function is triggered when a picture is uploaded to the storage bucket.

11. Explore the function

Let's have a closer look at the various interesting parts.

First, we're require-ing the needed modules, for Vision, Storage and Firestore:

const vision = require('@google-cloud/vision');
const Storage = require('@google-cloud/storage');
const Firestore = require('@google-cloud/firestore');

Then, we prepare a client for the Vision API:

const client = new vision.ImageAnnotatorClient();

Now comes the structure of our function. We make it an async function, as we are using the async / await capabilities introduced in Node.js 8:

exports.vision_analysis = async (event, context) => {
    ...
    const filename = event.name;
    const filebucket = event.bucket;
    ...
}

Notice the signature, but also how we retrieve the name of the file and bucket which triggered the Cloud Function.

For reference, here's what the event payload looks like:

{
  "bucket":"uploaded-pictures",
  "contentType":"image/png",
  "crc32c":"efhgyA==",
  "etag":"CKqB956MmucCEAE=",
  "generation":"1579795336773802",
  "id":"uploaded-pictures/Screenshot.png/1579795336773802",
  "kind":"storage#object",
  "md5Hash":"PN8Hukfrt6C7IyhZ8d3gfQ==",
  "mediaLink":"https://www.googleapis.com/download/storage/v1/b/uploaded-pictures/o/Screenshot.png?generation=1579795336773802&alt=media",
  "metageneration":"1",
  "name":"Screenshot.png",
  "selfLink":"https://www.googleapis.com/storage/v1/b/uploaded-pictures/o/Screenshot.png",
  "size":"173557",
  "storageClass":"STANDARD",
  "timeCreated":"2020-01-23T16:02:16.773Z",
  "timeStorageClassUpdated":"2020-01-23T16:02:16.773Z",
  "updated":"2020-01-23T16:02:16.773Z"
}

We prepare a request to send via the Vision client:

const request = {
    image: { source: { imageUri: `gs://${filebucket}/${filename}` } },
    features: [
        { type: 'LABEL_DETECTION' },
        { type: 'IMAGE_PROPERTIES' },
        { type: 'SAFE_SEARCH_DETECTION' }
    ]
};

We're asking for 3 key capabilities of the Vision API:

• Label detection: to understand what's in those pictures
• Image properties: to give interesting attributes of the picture (we're interested in the dominant color of the picture)
• Safe search: to know if the image is safe to show (it shouldn't contain adult / medical / racy / violent content)

At this point, we can make the call to the Vision API:

const [response] = await client.annotateImage(request);

For reference, here's what the response from the Vision API looks like:

{
  "faceAnnotations": [],
  "landmarkAnnotations": [],
  "logoAnnotations": [],
  "labelAnnotations": [
    {
      "locations": [],
      "properties": [],
      "mid": "/m/01yrx",
      "locale": "",
      "description": "Cat",
      "score": 0.9959855675697327,
      "confidence": 0,
      "topicality": 0.9959855675697327,
      "boundingPoly": null
    },
    ✄ - - - ✄
  ],
  "textAnnotations": [],
  "localizedObjectAnnotations": [],
  "safeSearchAnnotation": {
    "adult": "VERY_UNLIKELY",
    "spoof": "UNLIKELY",
    "medical": "VERY_UNLIKELY",
    "violence": "VERY_UNLIKELY",
    "racy": "VERY_UNLIKELY",
    "adultConfidence": 0,
    "spoofConfidence": 0,
    "medicalConfidence": 0,
    "violenceConfidence": 0,
    "racyConfidence": 0,
    "nsfwConfidence": 0
  },
  "imagePropertiesAnnotation": {
    "dominantColors": {
      "colors": [
        {
          "color": {
            "red": 203,
            "green": 201,
            "blue": 201,
            "alpha": null
          },
          "score": 0.4175916016101837,
          "pixelFraction": 0.44456374645233154
        },
        ✄ - - - ✄
      ]
    }
  },
  "error": null,
  "cropHintsAnnotation": {
    "cropHints": [
      {
        "boundingPoly": {
          "vertices": [
            { "x": 0, "y": 118 },
            { "x": 1177, "y": 118 },
            { "x": 1177, "y": 783 },
            { "x": 0, "y": 783 }
          ],
          "normalizedVertices": []
        },
        "confidence": 0.41695669293403625,
        "importanceFraction": 1
      }
    ]
  },
  "fullTextAnnotation": null,
  "webDetection": null,
  "productSearchResults": null,
  "context": null
}

If there's no error returned, we can move on, hence why we have this if block:

if (response.error === null) {
    ...
} else {
    throw new Error(`Vision API error: code ${response.error.code},  
                     message: "${response.error.message}"`);
}

We are going to get the labels of the things, categories or themes recognised in the picture:

const labels = response.labelAnnotations
    .sort((ann1, ann2) => ann2.score - ann1.score)
    .map(ann => ann.description)

We're sorting the labels by highest score first.

We're interested in knowing the dominant color of the picture:

const color = response.imagePropertiesAnnotation.dominantColors.colors
    .sort((c1, c2) => c2.score - c1.score)[0].color;
const colorHex = decColorToHex(color.red, color.green, color.blue);

We're again sorting colors by score and take the first one.

We're also using a utility function to transform the red / green / blue values into an hexadecimal color code that we can use in CSS stylesheets.

Let's check if the picture is safe to show:

const safeSearch = response.safeSearchAnnotation;
const isSafe = ["adult", "spoof", "medical", "violence", "racy"]
    .every(k => !['LIKELY', 'VERY_LIKELY'].includes(safeSearch[k]));

We're checking the adult / spoof / medical / violence / racy attributes to see if they are not likely or very likely to.

If the result of the safe search is okay, we can store metadata in Firestore:

if (isSafe) {
    const pictureStore = new Firestore().collection('pictures');
            
    const doc = pictureStore.doc(filename);
    await doc.set({
        labels: labels,
        color: colorHex,
        created: Firestore.Timestamp.now()
    }, {merge: true});
}

12. Deploy the function

Time to deploy the function.

Note: Make sure the Entry point text field below the code is set to vision_analysis so the correct JavaScript function is triggered when a picture is uploaded to the storage bucket. That's the name of your exported function in the JavaScript code (not necessarily the same name as the one of Cloud Function in the UI.

Hit DEPLOY button and the new version will be deployed, you can see the progress:

13. Test the function again

Once the function is successfully deployed, you will post a picture to Cloud Storage, see if our function is invoked, what the Vision API returns, and if metadata is stored in Firestore.

Navigate back to Cloud Storage, and click on the bucket we created at the beginning of the lab:

Once in the bucket details page, click on the Upload files button to upload a picture.

From the "hamburger" (☰) menu, navigate to the Logging > Logs Explorer.

In the Log Fields selector, select Cloud Function to see the logs dedicated to your functions. Scroll down through the Log Fields and you can even select a specific function to have a finer-grained view of the functions related logs. Select the picture-uploaded function.

And indeed, in the list of logs, I can see that our function was invoked:

The logs indicate the start and end of the function execution. And in between, we can see the logs we put in our function with the console.log() statements. We see:

• The details of the event triggering our function,
• The raw results from the Vision API call,
• The labels that were found in the picture we uploaded,
• The dominant colors information,
• Whether the picture is safe to show,
• And eventually those metadata about the picture have been stored in Firestore.

Again from the "hamburger" (☰) menu, go to the Firestore section. In the Data subsection (shown by default), you should see the pictures collection with a new document added, corresponding to the picture you just uploaded:

14. Clean up (Optional)

Note: If you intend to continue with other labs in the series, you can skip this step.

If you don't intend to continue with the other labs in the series, you can clean up resources to save costs and to be an overall good cloud citizen. You can clean up resources individually as follows.

Delete the bucket:

gsutil rb gs://${BUCKET_PICTURES}

Delete the function:

gcloud functions delete picture-uploaded --region europe-west1 -q

Delete the Firestore collection by selecting Delete collection from the collection:

Alternatively, you can delete the whole project:

gcloud projects delete ${GOOGLE_CLOUD_PROJECT} 

15. Congratulations!

Congratulations! You've successfully implemented the first key service of the project!

What we've covered

• Cloud Storage
• Cloud Functions
• Cloud Vision API
• Cloud Firestore

Next Steps

• Pic-a-daily: Lab 2—Creating thumbnails of big pictures