OpenCV 4 with Python Blueprints - Second Edition

Accuracy

The most straightforward metric to calculate is probably accuracy. This metric simply counts the number of test samples that have been predicted correctly, and returns the number as a fraction of the total number of test samples, as shown in the following code block:

def accuracy(y_predicted, y_true):    return sum(y_predicted == y_true) / len(y_true)

The previous code shows that we have extracted y_predicted by calling model.predict(x_test). This was quite simple, but, again, to make things reusable, we put this inside a function that takes predicted and true labels. And now, we will go on to implement slightly more complicated metrics that are useful to measure classifier performance.