Visual Recognition

In this lab, we'll create a custom classifier in IBM Watson Visual Recognition to recognize several dog breeds. We can then use this custom classifier to classify images and understand if there is certain breed of a dog in the image.

Getting Started

This lab uses the IBM Watson Visual Recognition. Sign up for an IBM Cloud account at ibm.biz/. This lab also uses the command line to train the Visual Recognition service with images. You can also use a programming language to programmatically train the service.

Create a Custom Classifier

Let's begin by creating an instance of the IBM Watson Visual Recognition service.

1. Click on the Catalog link in the top right corner of the IBM Cloud dashboard.

   

2. Click on the Watson category on the left. Click on the Visual Recognition tile.

   

3. Leave the service name as is and click Create.

   

4. Click on Service Credentials in the menu on the left. If there are no credentials in the list, click New credential and Add to create a set of credentials. Click on View Credentials to display the service credentials.

   
   

5. Copy the API Key. Run the command, replacing the API key with your own.

   

   export VR_KEY="g0h123kjf5h3m620n5h1175mrk54h32vc54ji543"
   

6. Run the command shown below to get a list of the classifiers.

   curl "https://gateway-a.watsonplatform.net/visual-recognition/api/v3/classifiers?api_key=$VR_KEY&version=2016-05-20"
   

   A sample response is shown below. No custom classifiers have been trained.

   {
       "classifiers": []
   }
   

7. Download the sample ZIP files. Each ZIP file contains a set of images with different breeds of dogs, and one has a set of images without any dogs. These images provide a basic training for IBM Watson to use to understand the differences between the classes we'll create in a moment.

   • https://watson-developer-cloud.github.io/doc-tutorial-downloads/visual-recognition/beagle.zip
   • https://watson-developer-cloud.github.io/doc-tutorial-downloads/visual-recognition/golden-retriever.zip
   • https://watson-developer-cloud.github.io/doc-tutorial-downloads/visual-recognition/husky.zip
   • https://watson-developer-cloud.github.io/doc-tutorial-downloads/visual-recognition/cats.zip

8. Create a custom classifier using these ZIP files.

   curl -X POST -F "beagle_positive_examples=@beagle.zip" -F "goldenretriever_positive_examples=@golden-retriever.zip" -F "husky_positive_examples=@husky.zip" -F "negative_examples=@cats.zip" -F "name=dogs" "https://gateway-a.watsonplatform.net/visual-recognition/api/v3/classifiers?api_key=$VR_KEY&version=2016-05-20"
   

9. Copy the classifier ID that is returned. In the example response below, the classifer ID is dogs_1992940714. Your classifier ID will be different.

   {
       "classifier_id": "dogs_1992940714",
       "name": "dogs",
       "status": "training",
       "owner": "b9baae18-2dd5-4517-ad68-7c9a651f3dab",
       "created": "2018-03-28T06:13:18.523Z",
       "updated": "2018-03-28T06:13:18.523Z",
       "classes": [
           {
               "class": "husky"
           },
           {
               "class": "goldenretriever"
           },
           {
               "class": "beagle"
           }
       ],
       "core_ml_enabled": true
   }
   

10. Run the command to save this classifier ID in a variable.

    export CLASSIFIER_ID="dogs_1992940714"
    

11. Check the status of the classifier by running the command shown below. It may take several minutes (or longer) for the training to complete.

    curl "https://gateway-a.watsonplatform.net/visual-recognition/api/v3/classifiers?api_key=$VR_KEY&version=2016-05-20"
    

    An example response is shown below.

    {
        "classifiers": [
            {
                "classifier_id": "dogs_1992940714",
                "name": "dogs",
                "status": "training"
            }
        ]
    }
    

    When the classifier has been completely trained, the status will become ready.

    {
        "classifiers": [
            {
                "classifier_id": "dogs_1992940714",
                "name": "dogs",
                "status": "ready"
            }
        ]
    }
    

12. When the status changes to ready, you can now classify images with this custom classifier.

    curl "https://gateway-a.watsonplatform.net/visual-recognition/api/v3/classify?api_key=$VR_KEY&url=http://cdn2-www.dogtime.com/assets/uploads/gallery/beagle-dog-breed-pictures/1-runforward.jpg&version=2016-05-20&classifier_ids=$CLASSIFIER_ID&threshold=0.0"
    

    An example response is shown below.

    {
        "images": [
            {
                "classifiers": [
                    {
                        "classifier_id": "dogs_1992940714",
                        "name": "dogs",
                        "classes": [
                            {
                                "class": "beagle",
                                "score": 0.888
                            },
                            {
                                "class": "goldenretriever",
                                "score": 0.02
                            },
                            {
                                "class": "husky",
                                "score": 0
                            }
                        ]
                    }
                ],
                "source_url": "http://cdn2-www.dogtime.com/assets/uploads/gallery/beagle-dog-breed-pictures/1-runforward.jpg",
                "resolved_url": "http://cdn2-www.dogtime.com/assets/uploads/gallery/beagle-dog-breed-pictures/1-runforward.jpg"
            }
        ],
        "images_processed": 1,
        "custom_classes": 3
    }
    

    Watson recognizes the class beagle.

13. Find another image on the internet and classify it by replacing the URL value in the previous step to point to the location of the image.