Keras: get labels name of pre-trained models on imagenet

Question

I am using Keras Inception_v3 pre-trained on Imagenet:

base_model = InceptionV3(weights='imagenet', include_top=True)

And when I predict from generated images, I get the output vector which has a shape (n,1000) with n being the number of images given. So now if I want to interpret the result, I need the name of the 1000 output classes used to train the model... But I can't find it!

Any idea ?

Answer 1

You can use decode_predictions method:

from keras.applications.inception_v3 import decode_predictions

preds = model.predict(x)
print('Predicted:', decode_predictions(preds, top=10))

# Predicted: [(u'n02504013', u'Indian_elephant', 0.0042589349), ...]

From source code:

def decode_predictions(preds, top=5, **kwargs):
    """Decodes the prediction of an ImageNet model.
    # Arguments
        preds: Numpy tensor encoding a batch of predictions.
        top: Integer, how many top-guesses to return.
    # Returns
        A list of lists of top class prediction tuples
        `(class_name, class_description, score)`.
        One list of tuples per sample in batch input.
    # Raises
        ValueError: In case of invalid shape of the `pred` array
            (must be 2D).
    """

Obviously it is not specific to Inception_V3. You can import it and use it for any pre-trained model on Imagenet. Alternatively, you can import it using:

from keras.applications.imagenet_utils import decode_predictions