Keras: Get true labels (y_test) from ImageDataGenerator or predict_generator
I am using ImageDataGenerator().flow_from_directory(...)
to create data packages from catalogs.
After successfully assembling the model, I would like to get a two-column array of True and Predicted class labels. With help model.predict_generator(validation_generator, steps=NUM_STEPS)
I can get many arrays of predicted classes. Is it possible to predict_generator
output the appropriate class labels True?
To add: validation_generator.classes does print the labels True, but in the order in which they are fetched from the directory, it does not account for batch expansion or fetch growth.
source to share
You can get the forecast labels:
y_pred = numpy.rint(predictions)
and you can get true shortcuts by:
y_true = validation_generator.classes
Before doing this, you must install shuffle=False
in the validation generator.
Finally, you can print the confusion matrix
print confusion_matrix(y_true, y_pred)
source to share