OCI Vision – Creating a Custom Model for Cats and Dogs
In this post, I’ll build on the previous work on preparing data, to using this dataset as input to building a Custom AI Vision model. In the previous post, the dataset was labelled into images containing Cats and Dogs. The following steps takes you through creating the Customer AI Vision model and to test this model using some different images of Cats.
Open the OCI Vision page. On the bottom left-hand side of the menu you will see Projects. Click on this to open the Projects page for creating a Custom AI Vision model.
On the Create Projects page, click on the Create Project button. A pop-up window will appear. Enter the name for the model and click on the Create Project bottom at the bottom of the pop-up.
After the Project has been created, click on the project name from the list. This will open the Project specific page. A project can contain multiple models and they will be listed here. For the Cats-and-Dogs project we and to create our model. Click on the Create Model button.
Next, you can define some of the settings for the Model. These include what dataset to use, or upload a new one, define what data labelling to use and the training duration. For this later setting, you can decide how much time you’d like to allocate to creating the custom model. Maybe consider selecting Quick mode, as that will give you a model within a few minutes (or up to an hour), whereas the other options can allow the model process to run for longer and hopefully those will create a more accurate model. As with all machine learning type models, you need to take some time to test which configuration works best for your scenario. In the following, the Quick mode option is selected. When read, click Create Model.
It can take a little bit of time to create the model. We selected the Quick mode, which has a maximum of one hour. In my scenario, the model build process was completed after four minutes. The Precentage Complete is updated during the build allowing you to monitor it’s progress.
When the model is completed, you can test it using the Model page. Just click on the link for the model and you’ll get a page like the one to the right.
The bottom half of this page allows you to upload and evaluate images. The following images are example images of cats (do you know the owner) and the predictions and information about these are displayed on the screen. Have a look at the following to see which images scored better than others for identifying a Cat.
Ora-lytics 