Once again, we can go to the Default Custom Model page, click on the Test tab, and drop some new test images for the model to evaluate and classify:

You can see that the model has correctly classified the first image as 0.77 angry. We also retested a previous image as a bit of a regression test and the model again correctly classified it (as 0.66 happy). 

Notice that now our model provides three scores for each test image: angry, happy, and sad, corresponding to each of our model's defined classes: