Testing the model

You trained model is now ready to integrate into a website, software applications, mobile phone apps, any way that you think you can use it.

How to use your model

In the Use your model section, there is sample code for starting your model, analyzing and image, and stopping your model.

Use your model

There are multiple methods and programming languages that you can use to do this so that you can integrate it into your business.

Starting and stopping a model

  • Select your model
  • Select Use Model
  • Select number of Inference units, choose 1 inference unit
  • Select Start Model

When your model is starting. It might take up to 30 minutes to start running.

See the full instructions for Starting or Stopping an Amazon Rekognition Custom Labels Model (Console)

There are some costs to training and running the model. Amazon Rekognition Custom Labels pricing. Search the pricing page for Pricing example 1 – Image Labeling for a Website

Amazon Rekognition Custom Labels Demo

Deploying the web application will be part of our next step and for now we will test a trained model that we deployed earlier.

You have done all of the hardwork creating a dataset, labelling your data, training a model, and evaluating the model results. Post in the chat how you would use your own custom label model in your own software.

Use a pretrained model

We have deployed a puppy model using the instructions above and deployed it with a Amazon Rekognition Custom Labels Demo web application above.

  1. Open a new web browser tab and go to https://dwizm1fya2bik.cloudfront.net/
  2. Login with the username puppy2

    puppy2
  3. Use password puppy2#

    puppy2$
  4. Choose the PuppyChallenge.2021-##-##T##.##.## model

  5. Upload an image of a puppy.

  6. View the results of the custom label.

Output of the model

You can see the Results from the train model output. The labels and confidence scores are listed and this web application draws a box around any puppies it has detected.

Look at the response to see the label applied to your image.

{
  "CustomLabels": [
    {
      "Confidence": 83.60199737548828,
      "Geometry": {
        "BoundingBox": {
          "Height": 0.6938999891281128,
          "Left": 0.2750700116157532,
          "Top": 0.23050999641418457,
          "Width": 0.4900299906730652
        }
      },
      "Name": "Puppy"
    }
  ]
}

Post in the chat if the model detected your puppy image.