Deploy Computer Vision Models with Wallaroo: Tutorial

Table of Contents

1. Introduction
2. Setting Up the Environment
3. Connecting to Wallaroo
4. Creating the Workspace
5. Provisioning the Pipeline
6. Uploading the Model
7. Deploying the Model
8. testing the Pipeline
9. Analyzing the Inference Results
10. Comparing and Monitoring Models
11. Conclusion

Introduction

In this Tutorial, we will learn how to use Wallaroo with computer vision models to detect objects in images. We will take you through the steps of transitioning your computer vision models from experimentation and development to production. Wallaroo is a platform that makes it easy to deploy models into production with just a few lines of code. This tutorial, along with many others, is available on our docs site. You can also run these tutorials in our free Community Edition, which can be installed on Azure, AWS, and GCP.

Setting Up the Environment

Before we get started with the tutorial, we need to set up our environment. We will be using Jupyter notebooks and the Wallaroo SDK to run our models and get them into production. To do this, we first need to import the necessary libraries. This is a common first step in many of our tutorials. Next, we will connect to Wallaroo. If you are not logged in, you may see a URL Prompt. Simply click on the URL and authenticate to proceed. Once connected, we will set up our variables including the workspace, pipeline, and model name. We will also specify the Onyx model format, but don't worry if your models are in a different format. Wallaroo provides tutorials on how to convert models into the Onyx format.

Connecting to Wallaroo

Now that our environment is set up, we can connect to Wallaroo. This will allow us to create the workspace and provision the pipeline. We will use the previously set variables to create the workspace and pipeline. Once completed, we are ready to move on to the next step.

Creating the Workspace

In order to deploy our model, we need to create a workspace. The workspace is where we will upload our model and set up the pipeline. We will create the workspace using the specified variables and verify that it has been created successfully.

Provisioning the Pipeline

With the workspace created, we can now provision the pipeline. The pipeline is the engine that will run our model in a production environment. Once the pipeline is provisioned, we are ready to upload our model and deploy it.

Uploading the Model

To upload our model, we will use the Wallaroo SDK. We will upload the model in the specified Onyx format. The model will then be available for deployment.

Deploying the Model

In this step, we will deploy the model into the production environment. What makes Wallaroo unique is its lightning-fast deployment speed. Unlike traditional methods that can take hours or even days, Wallaroo can deploy a model in under 45 seconds. This allows you to get your models into production quickly and efficiently.

Testing the Pipeline

With our model deployed, we can now test the pipeline. We will run some inference on a sample image and analyze the results. Before running the inference, we need to load, resize, and transform the image to meet the requirements of the ResNet image detector. Once the image is prepared, we can run the inference and see the results.

Analyzing the Inference Results

In this step, we will analyze the inference results. We will visually inspect the image and identify the objects detected by the model. In our example case, we are using a computer vision model to detect objects in a cashier-less store. We can see the objects and their confidence levels in the image. We can also view the listing of objects in a table format, which provides the confidence level of each item. This analysis is crucial in ensuring that the model is accurately detecting the objects it is trained for.

Comparing and Monitoring Models

Once the model is deployed, we can compare it with other models and monitor its performance over time. Wallaroo allows you to easily switch between different models and observe any model drift. This is important in ensuring that the models are performing as expected and making accurate predictions. We encourage you to explore our other videos and tutorials on model comparison and monitoring.

Conclusion

In this tutorial, we have learned how to use Wallaroo with computer vision models to detect objects in images. We have seen how easy it is to transition models from experimentation and development to production using the Wallaroo platform. With its lightning-fast deployment speed, Wallaroo allows you to get your models into production quickly and efficiently. We encourage you to try this tutorial yourself using our free Community Edition. Thank you for joining us, and we hope you found this tutorial helpful.

🌟 Highlights 🌟

• Transition your computer vision models from experimentation and development to production
• Use Wallaroo to deploy your models with just a few lines of code
• Upload and deploy models in the lightning-fast Wallaroo platform
• Test and analyze the inference results of your models
• Compare and monitor models to ensure accurate predictions

FAQ

Q: Can I use models in formats other than Onyx? A: Yes, Wallaroo provides tutorials on how to convert models into the Onyx format.

Q: How long does it take to deploy a model with Wallaroo? A: Wallaroo can deploy a model in under 45 seconds, which is significantly faster than traditional deployment methods.

Q: Can I compare and monitor models in Wallaroo? A: Yes, Wallaroo allows you to compare models and monitor their performance over time to ensure accurate predictions.