The idea is for our model to interpret the preceding images, perform some expression analysis, and then classify each image as either happy or sad facial expressions. In addition, the model should generate and display a score for each of the defined classes (except for the negative class).

As you have seen in our model, we defined just two classes to be classified—happy and sad. The model should, for each test image, display a percentage score showing the percentage of whether the detected expression is happy or sad. For example, the following score indicates that there is approximately 90 percent chance that the expression identified is happy: