How to find wrong prediction cases in test set (CN

问题:

I'm using MNIST example with 60000 training image and 10000 testing image. How do I find which of the 10000 testing image that has an incorrect classification/prediction?

回答1:

Simply use model.predict_classes() and compare the output with true labes. i.e:

incorrects = np.nonzero(model.predict_class(X_test).reshape((-1,)) != y_test)

to get indices of incorrect predictions

回答2:

To identify the image files that are wrongly classified, you can use:

imagenames = test_generator.filenames
errors = np.where(y_pred != test_generator.classes)[0]
for i in errors:
    print(fnames[i])