Create Custom Image Recognition Model with AI Builder
Table of Contents:
- Introduction to Power Automate AI Builder
- Setting Up a Model for Image Recognition
- Creating a New Model
- Defining the Objects to Detect
- Uploading Images for Training
- Tagging and Identifying Objects in Images
- Training the Model
- Publishing the Model
- Testing the Model
- Using the Model in Power Automate
Introduction to Power Automate AI Builder
Power Automate AI Builder is a powerful tool that allows You to Create and train your own models for image recognition. With AI Builder, you can easily detect and identify custom objects in images, opening up a world of possibilities for automation and data analysis. In this article, we will explore how to set up a model for image recognition using Power Automate AI Builder and walk through the step-by-step process of creating and training your own model.
Setting Up a Model for Image Recognition
To get started with creating a model for image recognition, you need to log into Power Automate and navigate to the AI Builder left menu. Under the "Models" section, you can create a new model. AI Builder provides a wizard that will guide you through the entire process, from defining the objects you want to detect to training the model and testing its accuracy.
Creating a New Model
In this particular case, we will be detecting custom objects in images. AI Builder allows you to define the objects you want to detect and provides predefined categories like common objects, retail shelves, and brand logos. For our example, we will select "common objects" and proceed to the next step.
Defining the Objects to Detect
The next step is to define the objects you want the model to detect. In our case, we will be identifying dice of different sizes used in tabletop gaming. We will provide at least 15 images for each object, ensuring a variety of angles, backgrounds, colors, and lighting conditions. The more examples you can provide, the better the accuracy of your model will be. AI Builder allows you to add images from your local device, SharePoint, or block storage.
Uploading Images for Training
Once you have defined the objects, you can add images for training. AI Builder recommends having a minimum of 15 images for each object but encourages providing at least 50 images for better accuracy. You can upload images from your local device, ensuring that they are in JPEG, PNG, or BMP format and not exceeding 6 megabytes in size. The images should showcase the objects from different perspectives and include variations in lighting and background.
Tagging and Identifying Objects in Images
After uploading the images, you will need to tag and identify the objects in each image. AI Builder provides a user-friendly interface where you can drag a box over the object and select the corresponding object label from a predefined list. You can also handle images with multiple objects by tagging each object separately. The more accurately you tag the objects in the images, the better the model's ability to detect them.
Training the Model
Once you have tagged and identified the objects in your images, you can proceed to the training step. Clicking on "Train" will initiate the model training process, which may take several minutes to several hours, depending on the size of your training dataset. AI Builder will provide status updates on the training progress, and once it is completed, you will have a trained model ready for use.
Publishing the Model
Before you can use the trained model in applications or workflows, you need to publish it. Publishing makes the model available for consumption by other processes and ensures that the latest version is used. Once the model is published, you can make changes and improvements to the model while working with it without affecting the version used in applications.
Testing the Model
To test the model, you can upload a new image and see if the model can correctly identify and detect the objects in the image. The accuracy of the model depends on the quality and diversity of the training data. AI Builder provides confidence scores for each identified object and allows you to extract the coordinates of the objects within the image.
Using the Model in Power Automate
Once you have a trained and published model, you can use it in Power Automate flows and PowerApps applications. Power Automate provides integration with AI Builder, allowing you to incorporate the model into your automation workflows. By using the "Use Model" action, you can pass an image to the model and receive the results, which can be further processed or used to trigger other actions.
In conclusion, Power Automate AI Builder offers a user-friendly and efficient way to create and train your own models for image recognition. By following the step-by-step process outlined in this article, you can define, train, and publish a model that accurately detects and identifies custom objects in images. With the power of AI Builder, you can automate tasks, gain valuable insights from image data, and enhance your business processes.