Printing tensor value output of model.output in keras

Question

So I'm trying to print the value of the model.output (which is a tensor) of pretrained VGG16 model, but unfortunately all the below listed methods didn't work in my case, which I found from the previously answered similar questions.

1) Firstly on the basis of this post I tried using

print(K.eval(model.output()))

but it threw an error as

TypeError: 'Tensor' object is not callable

2) Then upon going through this post I tried using

K.print_tensor(model.output, message='model.output = ')

approach but this time there was no output printed.

Here is my code:

from keras.applications.vgg16 import VGG16
from keras.applications.vgg16 import preprocess_input
from keras.preprocessing.image import load_img
from keras.preprocessing.image import img_to_array
from keras.models import Model
import keras.backend as K
from matplotlib import pyplot
from numpy import expand_dims
import numpy as np


# load the model
model = VGG16()

img = load_img('../input/treebird/bird.jpg', target_size=(224, 224))
img = img_to_array(img)           
img = expand_dims(img, axis=0)      
img = preprocess_input(img)       

prediction = model.predict(img)

print(model.output)                                 #Tensor("predictions_19/Softmax:0",       shape=(?, 1000), dtype=float32)
print(K.eval(model.output()))                          # throws TypeError: 'Tensor' object is not callable
K.print_tensor(model.output, message='model.output = ')

Am I lacking somewhere in the implementation of the above methods or are there other methods that I am supposed to use to print the tensor in this case?

score 2 · Accepted Answer · answered Sep 16 '19 at 19:03

Posting only the relevant parts of the code.

First you will have to clear the session so that your input node in model will have same name everytime. Use model.summary() to know what is the input node name.

Then, you will have to make use of feed_dict to pass the tensor as shown in the code.

K.clear_session()

# load the model
model = VGG16()
model.summary()

print(model.output.eval(session=K.get_session(), feed_dict={'input_1:0': img}))

Printing tensor value output of model.output in keras

1 Answers1