Another great improvement for our solution would be to add additional classes. This is to allow our model to support the detection of, perhaps, a third emotion or expression other than happy and sad. Let's try to add anger as our third (not counting negative) class:
- The first step, of course, is to collect and compress (zip up) our angry images training data:
Remember that it is not important what the individual image files are named, but it is important that they all represent the same emotion, anger.
- After we upload the angry.zip file (as another data asset available to our project), we can then go ahead and click on Create a class, enter angry for the class name, and then click on Create:
After a moment or two, our new angry class is ready:
- Now we can once more click on the Train Model button to start retraining the Face model on the training images we provided, along with our new angry class. After a few moments, we should see the Model Trained message again: