AWS Marketplace

How to try a product demo of a machine learning model from AWS Marketplace without subscribing to it

During my interactions, I hear feedback from builders on how they are looking to test drive machine learning (ML) models quickly. AWS Marketplace makes it easy to find, try, buy, and deploy ML models, which are deployed using Amazon SageMaker.

I’m pleased to announce that the Try Product demo feature is now available. This feature lets you perform a prediction on pre-trained ML models from AWS Marketplace without you having to subscribe and deploy them. See the launch tweet by Jeff Barr and a short demo video. This feature was recently enabled for the following model packages, and the AWS Marketplace team is continuously working on adding more to the list:

 Text Classification ML model packages:

Text generation ML Model Packages:

Named-Entity Recognition ML Model Packages:

 Object detection ML model packages:

Image classification ML model packages:

In this blog post, I will show you how to try an ML model from AWS Marketplace in matter of minutes.

How to try a product demo of a machine learning model from AWS Marketplace

Step 1: Explore AWS Marketplace and find a model to try

The first step is to explore AWS Marketplace for ML model packages and find the ML model you wish you to try.

You can explore these ML model packages available in AWS Marketplace.

AWS Marketplace contains hundreds of ML model packages from AWS and popular third-party sellers. I recommend you to do a search and find an ML model that solves an ML problem your business is interested in addressing. For example, I am trying to solve a business problem that requires a general categorization of video before applying domain-specific ML models. I want the model to identify if the video has a lot of people present outdoors. In that case, I would apply the activity detection ML model. Likewise, if the model detects that the video is indoors with furniture, I want to apply sentiment detection and interior item detection ML models.

To successfully choose which ML models to apply, the software does first level of analysis. To do so, I am using GluonCV YOLOv3 Object Detector ML model, as shown in the following screenshot.

This is a general-purpose model that is trained on the COCO dataset that can detect 80 common object categories.

Step 2: Perform a prediction on the model without subscribing to it

Many computer vision and Natural Language Processing (NLP) ML problems have a Try Product Demo feature enabled. This feature lets you perform a prediction on the model package without subscribing or deploying it.

  • To see if the model package offers a product demo, I choose the product to go its product detail page. In the GluonCV YOLOv3 Object Detector’s product detail page right sidebar, there is a Demo available box with a Try Product Demo button. This indicates that there is a product demo available for this product.
  • To perform a prediction, I choose the Try Product Demo
  • I then have the option to subscribe to the product, save it to a list, or try the product demo.
  • I choose Try Product Demo.
  • The following screenshot shows the product demo page for this product; it includes the headline Object detection followed by an Upload an image button.

  • I choose Upload an image and upload the following image of a worker in a warehouse driving a forklift full of products. Photo courtesy.

  • Within seconds, I get the response of three objects identified in the image: a truck, a person, and a dining table. Each object is listed in a row that identifies the object, offers a confidence rate, and gives the option to highlight the object in the image. Refer to the following screenshot.

In these results, I can hover over Highlight in image for each identified object to see the corresponding object highlighted in my image.

  • Further down the page, I can see the detailed model response. For each item, this response gives the pixel location from the right, left, bottom, and top, along with the item’s score and identification. Refer to the following screenshot.


This general-purpose ML model correctly identified a person and a table, and it identified a forklift more generally as a truck. If your ML problem statement is domain-specific, try domain-specific ML models. For example, the AKTE forklift detector, can identify a forklift.

Step 3: Subscribing and Deploying the ML Model (optional)

Once you are ready and happy with the ML Model, you can deploy it one of two ways:

  1. AWS Management Console
    1. For step-by-step walkthrough, follow Adding AI to your applications with ready-to-use models from AWS Marketplace blog post.
  2. Jupyter notebook
    1. For a tutorial, watch Deploy and Perform Inference on ML Models From AWS Marketplace Using a Jupyter Notebook.
    2. To try it out yourself using an Amazon SageMaker notebook instance, use this generic Jupyter notebook.

After deploying the ML model, you can perform a deep evaluation on your large real-world dataset to evaluate it for your production needs. If you need a customization to be made to the model to meet your specific requirements, reach out to for assistance.

Additional resources

Here are some additional resources I recommend checking out.


The Try Product Demo feature makes it faster for you to try an ML Model from AWS Marketplace. In this blog post I gave you an overview of how to try one of several ML models from AWS Marketplace.

About the author

kanchan waikarKanchan Waikar is a Senior Partner Solutions Architect at Amazon Web Services with AWS Marketplace for machine learning group. She has over 13 years of experience building, architecting, and managing natural language processing (NLP) and software development projects. She has a masters degree in computer science (data science major) and enjoys helping customers build solutions backed by AI/ML based AWS services and partner solutions.