web
You’re offline. This is a read only version of the page.
close
Skip to main content
Community site session details

Community site session details

Session Id :
Dynamics 365 Community / Blogs / Kylie Kiser / AI at the Zoo: Building Lob...

AI at the Zoo: Building Lobe AI Model

Views (0)
KylieKiser Profile Picture KylieKiser 2,174

Welcome back to the AI at the Zoo series, today we are talking through Building Lobe AI Model. In yesterday’s introduction post, we downloaded Lobe and set up our trial environment. The next step will be to launch Lobe and build a new model.

Start Lobe

Lobe AI Welcome screen
After downloading and launching Lobe there will be a few welcome screens and information pages.
Lobe Create new Model
Next you can create a new model.

When creating a new model, you can click on the “Untitled” area to give it a name. Then click on Import to start adding your photos.

Import and Tag Images

Lobe Imported Photos. Multiple pictures of animals. No labels added.
After importing the photos, you will see them all as Unlabeled. Start updating each one with a label with what the image is of.

You will need a minimum of 5 images per label. Also try to get a variety of images so that the model can recognize different angles and characteristics. I just searched for images online to find a good selection for training.

Lobe Image Labeling
You will see the labels start suggesting existing labels as you type. Count of each label is visible on the left. When you reach 5 per label the training will automatically start.

Train the AI Model

Lobe Starting to Train the Model
When 5 images have been added for each label, the training will automatically start. This will go pretty quickly for simple models.
Lobe Training Complete
When training is complete, click on the Train area to see how it did predicting the labels for the existing images. The model will continue to be refined as more images are added and then is continually run against the training set to determine accuracy.

Test the Model

Testing the Lobe AI Model. Panda correctly identified as a Panda.
Now, click on the Use area and you can drop in additional images to test. It will tell you the expected label and you can indicate if it is correct or not. This will then be added into the model so it can be further refined.

Now, your model is ready to use! As you can see, these models can be created in just a few minutes! Then you can keep refining is as you get more data. Tomorrow we will look at how to export this model to the Power Platform.

Watch for more posts in the AI at the Zoo series or check out Events where you can find me presenting on AI or other Power Platform topics!

The post AI at the Zoo: Building Lobe AI Model appeared first on Kylie Kiser.


This was originally posted here.

Comments

*This post is locked for comments