#AzureML – Using a custom images 🖼️ dataset on the AzureML Designer with the DenseNet template 🎨

Hi !

In my previous post I wrote about how to create a custom dataset with images to be used on a Azure Machine Learning Designer project. Today I’ll share the steps to use it.

Let’s start by creating a project based on the “Image Classification using DenseNet” template.

new designer project based on the DenseNet template

Next step is to select the compute instance we want to use in this Pipeline

select compute instance for experiment

Let’s delete the original animal datasource and add the newly created squirrel dataset.

Connect the new dataset with the [Convert to Image Directory] asset. To double check that we have the right contents, right-click on the Dataset asset and select [Preview Data], a list with the first 50 elements must appear on the preview window.

connect the squirrel dataset with the 1st step and preview the dataset

Now we can run the model. This may take a couple of minutes, so time for check twitter 😀

running pipeline to train the model

Once the pipeline run is done, we can check the score and metrics of the trained model. And, this is important for my online sessions, we can also download the trained model. Click on the [Train PyTorch Model] asset, select [Output and Logs] in the preview window, and we can download the file!

download the generated model

Tomorrow I’ll show how to create a HTTP Rest Endpoint to consume the generated model.

Happy coding!

Greetings

El Bruno

More posts in my blog ElBruno.com.



¿Con ganas de ponerte al día?

En Lemoncode te ofrecemos formación online impartida por profesionales que se baten el cobre en consultoría:

Leave a comment

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: