How to Create an Image Model

What Is Machine Learning 2.0

First, what is machine learning? Imagine that your computer can learn new skills by observing and analyzing, such as recognizing animals or understanding simple instructions. This is what we call machine learning.

With machine learning, machines become intelligent. Now, we have integrated this kind of intelligence into mBlock to enable the learning and thinking abilities in your project.

With Machine Learning 2.0, even an AI novice can build an AI model in a few simple steps.

How to Create an Image Model

In this video, we will create an image model, including the mBot2 and panda classes, to show you how to fulfill an image classification task using Machine Learning 2.0.

First, click the Sprites tab, click extension, and add Machine Learning 2.0.

Then, click Create/Manage model to create a machine learning task.

On the New project page, click New model under Image Project and name your model.

On the page for model training, you can add data on the left panel for your classification task.

First, name the first class "mBot2", click Webcam, and collect several images.

Then, name the second class "panda", click Webcam, and collect several panda images. To add more objects for this model, click Add a class.

After collecting the data, click Train Model in the Training area. mBlock automatically learns how to recognize mBot2 and panda based on the image data. To modify the training parameters, click Advanced.

After completing model training, you can preview the classification results on the right panel. The probability that an object recognized belongs to each class is displayed under the video panel.

You can modify the dataset or parameters anytime to retrain the model. If you are satisfied with the classification result, click Use model in the upper-right corner to integrate the model into blocks. Now you can apply these AI-based blocks to your mBlock projects.

Was this article helpful?
1 out of 1 found this helpful

Comments

0 comments

Please sign in to leave a comment.