Hot Dog or Not Hot Dog Using Machine Learning

In this chapter, we will learn how to use machine learning to create a model that we can use for image classification. We will export the model as an Onnx model that we can use on all platforms – that is, Android, iOS, macOS, and Windows. To train and export models, we will use Azure Cognitive Services and the Custom Vision service.

Once we have exported the models, we will learn how to use them in a .NET MAUI app.

The following topics will be covered in this chapter:

• Training a model with Azure Cognitive Services and the Custom Vision service
• Using Onnx models for image classification using ML.NET
• Using custom routes in .NET MAUI for navigation

Technical requirements

To be able to complete this project, you need to have Visual Studio for Mac or PC installed, as well as the .NET MAUI components. See Chapter 1, Introduction to .NET MAUI, for more details on how to set up your environment. You also need an Azure account. If you have a Visual Studio subscription, there are a specific amount of Azure credits included each month. To activate your Azure benefits, go to https://my.visualstudio.com.

You can also create a free account, where you can use selected services for free over 12 months. You will get $200 worth of credit to explore any Azure service for 30 days, and you can also use the free services at any time. Read more at https://azure.microsoft.com/en-us/free/.

If you do not have and do not want to sign up for a free Azure account, the trained model is available in the source code for this chapter. You can download and use the pre-trained model instead.

The source code for this chapter is available at the GitHub...

Machine learning

The term machine learning was coined in 1959 by Arthur Samuel, an American pioneer in artificial intelligence (AI). Tom M. Mitchell, an American computer scientist, provided the following more formal definition of machine learning later:

“A computer program is said to learn from experience E with respect to some class of tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E.”

In simpler terms, this quote describes a computer program that can learn without being explicitly programmed. In machine learning, algorithms are used to build a mathematical model of sample data or training data. The models are used for computer programs to make predictions and decisions without being explicitly programmed for the task in question.

In this section, we will learn about a few different machine learning services and APIs that are available when developing a .NET MAUI application. Some APIs are only...

The project overview

If you have seen the TV series Silicon Valley, you have probably heard of the Not Hotdog application. In this chapter, we will learn how to build that app. The first part of this chapter will involve collecting the data that we will use to create a machine learning model that can detect whether a photo contains a hot dog.

In the second part of the chapter, we will build an app using .NET MAUI and ML.NET, where the user can either take a new photo or pick a photo in the photo library, analyzing it to see whether it contains a hot dog. The estimated time for completing this project is 120 minutes.

Getting started

We can use either Visual Studio 2022 on a PC or Visual Studio for Mac to do this project. To build an iOS app using Visual Studio for PC, you must have a Mac connected. If you don’t have access to a Mac at all, you can choose to just do the Android and Windows parts of this project.

Similarly, if you only have a Mac, you can choose to just do the iOS and macOS or Android parts of this project.

Building the Hot Dog or Not Hot Dog application using machine learning

Let’s get started! We will first train a model for image classification that we can use later in the chapter to decide whether a photo contains a hot dog.

Note

If you do not want to go through the effort of training a model, you can download a pre-trained model from the following URL: https://github.com/PacktPublishing/MAUI-Projects-3rd-Edition/tree/main/Chapter12/HotdogOrNot/Resources/Raw.

Training a model

To train a model for image classification, we need to collect photos of hot dogs and photos that aren’t of hot dogs. Because most items in the world are not hot dogs, we need more photos that don’t contain hot dogs. It’s better if the photos of hot dogs cover a lot of different hot dog scenarios – with bread, ketchup, or mustard. This is so that the model will be able to recognize hot dogs in different situations. When we collect photos that aren’t of hot...

Summary

In this chapter, we built an app that can recognize whether a photo contains a hot dog or not. We accomplished this by training a machine learning model for image classification, using Azure Cognitive Services and the Custom Vision service.

We exported models for ML.NET, and we learned how to use it in an MAUI app that targets iOS, Mac Catalyst, Windows, and Android. In the app, a user can take a photo or pick one from their photo library. This photo will be sent to the model to be classified, and we will get a result that tells us whether the photo is of a hot dog.

Now, we can continue to build other apps and use what we have learned in this chapter regarding machine learning, both on-device and in the cloud using Azure Cognitive Services. Even if we are building other apps, the concept will be the same.

Now, we have completed all the chapters in this book. We have learned the following:

• What .NET MAUI is and how we can get started building apps
• How to...