Skip to content

Latest commit

 

History

History
710 lines (551 loc) · 25.3 KB

README.md

File metadata and controls

710 lines (551 loc) · 25.3 KB
Raw

gablogo

Lab 5 - Azure Cognitive Services

Goal

Our goal is to know if a picture has a dog or not.

To do this, we will add Azure Cognitive Services (Vision) to our project. The web application will be modified to read one image (provided) from Azure blob storage and display the information returned by the Vision API. The image (a dog or a cat) will be uploaded to the Blob storage using the Azure Portal or the Azure Storage Explorer.

Let's code!

Add Azure components

Let's add Cognitive Services to our deployment (ARM) template and deploy it.

Modify the ARM template to add the Computer Vision service

In Lab 2 we created an ARM template to deploy the backbone of our application (app service plan, web app (mvc) and storage account).

We will start from that template and add the Cognitive Services resource to it

Open the existing ARM template and ARM template parameters from lab2.

1 - Add a 'csVisionName' variable to name the Cognitive Services resource

We will create a unique name for the vision resources utilizing the 'suffix' variable created previously.

The variable section should now look like this:

"variables": {
    "suffix": "[uniqueString(resourceGroup().id, resourceGroup().location)]",
    "storageName": "[concat('stg', variables('suffix'))]",
    "csVisionName": "[concat('vision', variables('suffix'))]"
}

2 - Add the Cognitive Services resource:

Firstly, under the resources array in the ARM template (gab2019.json) add:

{
    "type": "Microsoft.CognitiveServices/accounts",
    "apiVersion": "2016-02-01-preview",
    "name": "[variables('csVisionName')]",
    "location": "[resourceGroup().location]",
    "sku": {
        "name": "F0"
    },
    "kind": "ComputerVision",
    "dependsOn": [],
    "properties": {},
    "scale": null
}

This will tell Azure that we want an instance of Cognitive Services.

3 - Add the Cognitive Services access keys in the web application configuration

In the web app resource / parameters section of the ARM template (gab2019.json), add:

"properties": {
    "name": "[parameters('webAppName')]",
    "siteConfig": {
        "appSettings": [
        {
           "name": "ComputerVision:ApiEndpoint",
           "value": "[reference(variables('csVisionName'), '2017-04-18').endpoint]"
        },
        {
           "name": "ComputerVision:ApiKey",
           "value": "[listKeys(variables('csVisionName'), '2017-04-18').key1]"
        }]
    },
    "serverFarmId": "[resourceId('Microsoft.Web/serverfarms/', parameters('appSvcPlanName'))]"
},

The site config will add app settings to the web application in Azure.

The Cognitive Services endpoint and key will be set into our MVC Web App (just like we do with the storage account)

4 - Add output variables

In the output section of the template, we will add outputs to make it easier to get the keys and the connection string we will need to run our app locally. Replace the outputs section with the code below:

"outputs": {
    "CognitiveServices-endpoint": {
        "type": "string",
        "value": "[reference(variables('csVisionName'), '2017-04-18').endpoint]"
    },
    "CognitiveServices-key1": {
        "type": "string",
        "value": "[listKeys(variables('csVisionName'), '2017-04-18').key1]"
    },
    "Storage-connectionString": {
        "type": "string",
        "value": "[Concat('DefaultEndpointsProtocol=https;AccountName=',variables('storageName'),';AccountKey=',listKeys(resourceId('Microsoft.Storage/storageAccounts', variables('storageName')), providers('Microsoft.Storage', 'storageAccounts').apiVersions[0]).keys[0].value)]"
    }
}

This will add output variables to the resource group deployment.

At this point the ARM template for lab 5 is ready.

Deploy the ARM templates to Azure

Now that we have an ARM template that includes Cognitive Services, we want to deploy it.

In Lab 2 we saw a few methods to deploy the initial components of the ARM template. We will thus use the CI/CD approach to deploy our new resource.

Commit your changes to your git repository. Once that is done, the CI/CD will take over and deploy the new resource.

Getting config info for the web app

The ARM template deployment takes care of setting the config information in the web app, so there is no need to deploy the values in the app settings (or to commit these values to git).

We, however, will need the config information to run the web application locally. Copy these values into a text editor once the ARM templace has finished deploying or follow the steps below to get them from the portal. Refer yourself to Annex A to learn other ways to get all this information.

Getting config info from deployment

Since we added output variables to the ARM template, it makes it easier to get the information we need for Cognitive Services and Azure Storage.

To find the output values;

  • navigate to the Azure portal
  • locate your resource group
  • in the deployments section
    • click on the latest deployment
      • click on the outputs section
      • the output variable values will be there ready to be used (all in one place!)

deployment-output

Upload the dog image into your storage account

We will upload an image to blob storage to serve as test data.

  • create a container in your storage blob called images
  • upload the dog image found in the data directory into this container
    • Use either the portal or the storage explorer application (see Lab 0 for details)

Make sure to set the access level to public so the computer vision can read from it

To do that in Azure Storage Explorer;

  • right click on your container
  • select Set Public Access Level

public-access-blob-se

To do the same in the Azure Portal;

  • navigate to your storage account
  • click on the Blobs under Blob service
  • select your blob (in our case images)
  • In the overview blade, click the Change access level
  • Select Blob (anonymous read access for blobs only) as the public access level

public-access-blob-portal-overview

Let's add computer vision to our web application

Azure components are now deployed and test data is ready. It s now time to add Computer Vision functionality to the web application.

Adding the Computer Vision API library to the application

We need to install the Computer Vision client library NuGet package.

To do so;

  • open the terminal in your Visual Studio Code (ctrl+`)
  • type dotnet add package Microsoft.Azure.CognitiveServices.Vision.ComputerVision

Adding the Blob storage library to the application

We need to install the Blob storage library NuGet Package.

To do so;

  • again in your terminal
  • type dotnet add package Microsoft.Azure.Storage.Blob

Using the API in code

Setting up the resources keys

To use the computer vision API, first copy the API key you copied earlier into the file appsettings.json, under the ComputerVision:ApiKey section.

In the ComputerVision:ApiEndPoint section, enter the computer vision API endpoint. It should like like this: https://eastus.api.cognitive.microsoft.com, where eastus is the region into which you created your computer vision API resource into.

Now for the storage connection string, copy the connection string you copied earlier into the section ConnectionStrings:ApplicationStorage.

Note: The Key1:Key2, for instance ComputerVision:ApiEndPoint, refers to ASP.NET Core way of setting application settings in appsettings.json. The Key1:Key2 refers to the hierarchy of JSON keys. For instance, your appsettings.json should look like this:

{
  "Logging": {
    "LogLevel": {
      "Default": "Warning"
    }
  },
  "AllowedHosts": "*",
  "ConnectionStrings": {
    "ApplicationStorage": ""
  },
  "ComputerVision": {
    "ApiKey": "",
    "ApiEndPoint": ""
  }
}

Add code to make use of the resource keys

To be able to use these keys into our application, create two classes in the application named StorageAccountOptions and ComputerVisionOptions. We will be using the Options pattern available in ASP.NET Core. In that class, you will need to copy the structure of your appsettings into code. It should look like the following:

public class StorageAccountOptions
{
    [Required]
    public string ConnectionString { get; set; }
}
public class ComputerVisionOptions
{
    /// <summary>
    /// Your subscription key
    /// </summary>
    [Required]
    public string ApiKey { get; set; }

    /// <summary>
    /// The endpoint of the region in which your created your ComputerVision resource. i.e. https://westcentralus.api.cognitive.microsoft.com
    /// </summary>
    [Required]
    public string ApiEndPoint { get; set; }
}

Note that we have also created classes for the sub-sections.

Creating classes to consume the storage and computer vision API

Under the root of your application, create a folder called Services and create 2 classes (2 classes): BlobStorageManager and ImageAnalyzer

The BlobStorageManager is used to consume the blob storage and the ImageAnalyzer is used to consume the computer vision API.

BlobStorageManager

In order to consume the blob storage, copy the following code into the class BlobStorageManager

private readonly CloudStorageAccount _storageAccount;

public BlobStorageManager(IOptions<StorageAccountOptions> options)
{
    if (options == null) { throw new ArgumentNullException(nameof(options)); }
    _storageAccount = CreateCloudStorageAccount(options.Value);
}

private CloudStorageAccount CreateCloudStorageAccount(StorageAccountOptions options)
{
    if (!CloudStorageAccount.TryParse(options.ConnectionString, out CloudStorageAccount storageAccount))
    {
        throw new Exception("Invalid storage account connecting string. Please verify the connection string and try again");
    }
    return storageAccount;
}

public IEnumerable<IListBlobItem> GetFiles(string containerName)
{
    var cloudBlobClient = _storageAccount.CreateCloudBlobClient();
    var container = cloudBlobClient.GetContainerReference(containerName);
    var blobs = container.ListBlobs();
    return blobs;
}

The method GetFiles will be used to list all the files in your blob container (in our cases we will use the images container created earlier).

The namespaces for this class should have:

using System;
using System.Collections.Generic;
using Microsoft.Azure.Storage;
using Microsoft.Azure.Storage.Blob;
using Microsoft.Extensions.Options;

Create an interface that will be used for the dependency injection called IBlobStorageManager. Copy the following code into this interface

IEnumerable<IListBlobItem> GetFiles(string containerName);

Make sure to inherit from this interface. Your class definition should look like

public class BlobStorageManager : IBlobStorageManager

ImageAnalyzer

To consume the computer vision API, copy the following code into the class ImageAnalyzer

private readonly ComputerVisionClient _computerVision;

private static readonly List<VisualFeatureTypes> Features =
    new List<VisualFeatureTypes>
    {
        VisualFeatureTypes.Categories, VisualFeatureTypes.Description,
        VisualFeatureTypes.Faces, VisualFeatureTypes.ImageType,
        VisualFeatureTypes.Tags
    };

public ImageAnalyzer(ComputerVisionClient computerVision)
{
    _computerVision = computerVision ?? throw new ArgumentNullException(nameof(computerVision));
}

public Task<ImageAnalysis> AnalyzeAsync(string imageUrl)
{
    if (!Uri.IsWellFormedUriString(imageUrl, UriKind.Absolute))
    {
        throw new Exception($"Invalid remoteImageUrl: {imageUrl}");
    }
    var analysisResults = _computerVision.AnalyzeImageAsync(imageUrl, Features);
    return analysisResults;
}

As you can see, we are asking the API to return use the Categories of the image, its description, the image type information, the image tags and the image faces collection. You can also ask it to return the Color information (by adding VisualFeatureTypes.Color) and Objects information (by adding VisualFeatureTypes.Objects).

The namespaces for this class should have:

using System;
using System.Collections.Generic;
using System.Threading.Tasks;
using Microsoft.Azure.CognitiveServices.Vision.ComputerVision;
using Microsoft.Azure.CognitiveServices.Vision.ComputerVision.Models;

Create an interface that will be used for the dependency injection called IImageAnalyzer. Copy the following code into this interface

Task<ImageAnalysis> AnalyzeAsync(string imageUrl);

Make sure to inherit from this interface. Your class definition should look like

public class ImageAnalyzer : IImageAnalyzer

Wiring all this into the Startup

In your Startup class, under the ConfigureServices method, add the following:

// Blob Storage
services.AddOptions<StorageAccountOptions>()
    .Configure(options => options.ConnectionString = Configuration.GetConnectionString("ApplicationStorage"))
    .ValidateDataAnnotations();
services.AddSingleton<IBlobStorageManager, BlobStorageManager>();

// Computer Vision
services.AddOptions<ComputerVisionOptions>()
    .Bind(Configuration.GetSection("ComputerVision"))
    .ValidateDataAnnotations();
services.AddSingleton(serviceProvider =>
{
    var options = serviceProvider.GetRequiredService<IOptions<ComputerVisionOptions>>().Value;
    return new ComputerVisionClient(new ApiKeyServiceClientCredentials(options.ApiKey)) { Endpoint = options.ApiEndPoint };
});
services.AddSingleton<IImageAnalyzer, ImageAnalyzer>();

This will map our settings to our Options objects and add our service class into our dependency injection container for later consumption.

Make sure to add the following the following namespace to your startup class:

using Microsoft.Azure.CognitiveServices.Vision.ComputerVision;

Creating the controller and view

In the controller folder, create a class named AnalyzerController. This is the controller that will be called by our application to deal with the image analysis.

In that class, copy the following code:

private readonly IBlobStorageManager _blobStorageManager;
private readonly IImageAnalyzer _imageAnalyzer;

public AnalyzerController(IBlobStorageManager blobStorageManager, IImageAnalyzer imageAnalyzer)
{
    _blobStorageManager = blobStorageManager ?? throw new ArgumentNullException(nameof(blobStorageManager));
    _imageAnalyzer = imageAnalyzer ?? throw new ArgumentNullException(nameof(imageAnalyzer));
}

public IActionResult Index()
{
    var files = _blobStorageManager.GetFiles("images").Select(item => item.Uri).ToList();
    ViewBag.Files = files;
    return View();
}

public async Task<IActionResult> Analyze(string imageUrl)
{
    var results = await _imageAnalyzer.AnalyzeAsync(imageUrl);
    ViewData["Title"] = "Image analysis results";
    ViewData["ImageUrl"] = imageUrl;
    return View("Results", results);
}

The Index method will list all the images from our images container and the Analyze method will take the URL of our image in our container and analyze it with the computer vision API.

The namespaces in this class should include:

using System;
using System.Linq;
using System.Threading.Tasks;
using Microsoft.AspNetCore.Mvc;
using GABDemo.Services;

In the Views folder, create a new folder called Analyzer. In that folder add 2 Razor Views called Index and Results.

In the Index view, copy the following code:

<table class="table">
    <thead>
        <tr>
            <th scope="col">#</th>
            <th scope="col">Image</th>
            <th scope="col">Actions</th>
        </tr>
    </thead>
    <tbody>
        @if (ViewBag.Files != null) { int count = 1; foreach (Uri file in ViewBag.Files) {
        <tr>
            <th scope="row">@count</th>
            <td>@file</td>
            <td>
                <a asp-action="Analyze" asp-route-imageUrl="@file.AbsoluteUri">Analyze</a>
            </td>
        </tr>
        count++; } }
    </tbody>
</table>

This code displays all the files in our images container.

In the Results view, the most important part is having the model defined. At the top of the file, add the following:

@model Microsoft.Azure.CognitiveServices.Vision.ComputerVision.Models.ImageAnalysis

With this, you can access Model in your view and display the results information the way you want it. Here's an example of how it can be displayed:

<h1>Categories</h1>
<table class="table">
    <thead>
        <tr>
            <th scope="col">Name</th>
            <th scope="col">Score</th>
            <th scope="col">Details</th>
        </tr>
    </thead>
    <tbody>
        @foreach (var category in Model.Categories) {
        <tr>
            <td>@category.Name</td>
            <td>@category.Score</td>
            <td>
                @if (category.Detail != null) { if (category.Detail.Celebrities != null && category.Detail.Celebrities.Any()) { foreach (var
                celebrity in category.Detail.Celebrities) { @celebrity<br />
                } } if (category.Detail.Landmarks != null && category.Detail.Landmarks.Any()) { foreach (var landmark in
                category.Detail.Landmarks) { @landmark<br />
                } } }
            </td>
        </tr>
        }
    </tbody>
</table>

<h1>Description</h1>
<h2>Captions</h2>
<table class="table">
    <thead>
        <tr>
            <th scope="col">Text</th>
            <th scope="col">Confidence</th>
        </tr>
    </thead>
    @foreach (var caption in Model.Description.Captions) {
    <tr>
        <td>@caption.Text</td>
        <td>@caption.Confidence</td>
    </tr>
    }
</table>

<h2>Tags</h2>
@string.Join(", ", Model.Description.Tags) @if (Model.Adult != null) {
<h2>Adult content</h2>
<table class="table">
    <thead>
        <tr>
            <th scope="col">Adult score</th>
            <th scope="col">Is adult content?</th>
            <th scope="col">Is racy content?</th>
            <th scope="col">Racy score</th>
        </tr>
    </thead>
    <tbody>
        <tr>
            <td>@Model.Adult.AdultScore</td>
            <td>@Model.Adult.IsAdultContent</td>
            <td>@Model.Adult.IsRacyContent</td>
            <td>@Model.Adult.RacyScore</td>
        </tr>
    </tbody>
</table>
}

<h2>Image Type</h2>
Clipart confidence level: @Model.ImageType.ClipArtType<br />
Line drawing confidence level: @Model.ImageType.LineDrawingType<br />

@if (Model.Faces != null && Model.Faces.Count > 0) {
<h2>Faces</h2>
<table class="table">
    <thead>
        <tr>
            <th scope="col">Age</th>
            <th scope="col">Gender (if applicable)</th>
            <th scope="col">Face rectangle</th>
        </tr>
    </thead>
    <tbody>
        @foreach (var faceDescription in @Model.Faces) {
        <tr>
            <td>@faceDescription.Age</td>
            <td>@(faceDescription.Gender != null ? faceDescription.Gender.Value.ToString() : "N/A")</td>
            <td>
                Height: @faceDescription.FaceRectangle.Height<br />
                Width: @faceDescription.FaceRectangle.Width<br />
                Left: @faceDescription.FaceRectangle.Left<br />
                Top: @faceDescription.FaceRectangle.Width<br />
            </td>
        </tr>
        }
    </tbody>
</table>
}

<div class="row">
    <div class="col-sm">
        <h2>Image Metadata</h2>
        Width: @Model.Metadata.Width<br />
        Height: @Model.Metadata.Height<br />
        Format: @Model.Metadata.Format<br />
    </div>
    <div class="col-sm">
        <h2>Image</h2>
        <img src="@ViewData["ImageUrl"]" class="img-thumbnail" />
    </div>
</div>

Better options validation

If your application throws an OptionsValidationException and that you want a better understanding of the error, add a class named OptionsValidationExceptionFilterAttribute at the root of your project and replace the class by the following code:

/// <summary>
/// This filter allows for displaying more explicit options validation exception.
/// Implements the <see cref="Microsoft.AspNetCore.Mvc.Filters.ExceptionFilterAttribute" />
/// </summary>
/// <seealso cref="Microsoft.AspNetCore.Mvc.Filters.ExceptionFilterAttribute" />
public class OptionsValidationExceptionFilterAttribute : ExceptionFilterAttribute
{
    /// <summary>
    /// This method is called when an exception occurs and wrap <see cref="Microsoft.Extensions.Options.OptionsValidationException"/>
    /// into another <see cref="Exception"/> to make the message easier to understand.
    /// </summary>
    /// <param name="context">The provided exception context.</param>
    /// <inheritdoc />
    public override void OnException(ExceptionContext context)
    {
        if (context.Exception is OptionsValidationException validationEx)
        {
            context.Exception = new Exception(validationEx.Failures.First(), validationEx);
        }
    }
}

Then, in Startup.cs, replace .AddMvc() by .AddMvc(options => options.Filters.Add<OptionsValidationExceptionFilterAttribute>()).

Re-run your application and the error message should be clearer. If it is a generic error message, make sure that the ASPNETCORE_ENVIRONMENT is set to Development.

Folder structure

If you followed the steps above, your folder structure should look like this

folder-structure

Annex

A

Getting the API keys for the API

To get your computer vision keys;

  • navigate to the Azure portal
  • locate your computer vision resource
  • go into the keys section
    • find the key value under key1
    • click the Copy button to copy the key

computer-vision-keys

Getting the blob storage connection string

To get your storage key;

  • navigate to the Azure portal
  • locate your storage account
  • in the Settings section of the storage account overview
    • select Access keys
      • here you can view your account access keys and the complete connection string for each key
    • find the Connection string value under key1
    • click the Copy button to copy the connection string

blob-connectionstring

Reference

Quickstart: Analyze an image using the Computer Vision SDK and C#

Develop with blobs

Options pattern in ASP.NET Core

End

Previous Lab | Next Lab