Scipy.ndimage.interpolation raises exception on rotate in utils.input_modifiers

[fabianostermann]

Hi there, thank you for this great library!

I've stepped over a problem on trying to "Visualizing Dense layer using ActivationMaximization" as in your perfect example.

I updated everything and the example worked as expected. When I run with my own fully dense model I get following error (model summary in front, exact same for the other example with conv!):

Model: "sequential_4"
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
flatten_4 (Flatten)          (None, 128000)            0         
_________________________________________________________________
dense_16 (Dense)             (None, 256)               32768256  
_________________________________________________________________
dropout_12 (Dropout)         (None, 256)               0         
_________________________________________________________________
dense_17 (Dense)             (None, 128)               32896     
_________________________________________________________________
dropout_13 (Dropout)         (None, 128)               0         
_________________________________________________________________
dense_18 (Dense)             (None, 64)                8256      
_________________________________________________________________
dropout_14 (Dropout)         (None, 64)                0         
_________________________________________________________________
dense_19 (Dense)             (None, 2)                 130       
=================================================================
Total params: 32,809,538
Trainable params: 32,809,538
Non-trainable params: 0
_________________________________________________________________
Traceback (most recent call last):
  File "explain_dense.py", line 47, in <module>
    activations = activation_maximization(score, steps=256, callbacks=[PrintLogger(interval=50)])
  File "/home/oyster/.local/lib/python3.8/site-packages/tf_keras_vis/activation_maximization/__init__.py", line 111, in __call__
    seed_inputs[j] = modifier(seed_inputs[j])
  File "/home/oyster/.local/lib/python3.8/site-packages/tf_keras_vis/utils/input_modifiers.py", line 56, in __call__
    seed_input = rotate(seed_input,
  File "/home/oyster/.local/lib/python3.8/site-packages/scipy/ndimage/interpolation.py", line 686, in rotate
    raise ValueError('axes should contain exactly two values')
ValueError: axes should contain exactly two values

I debugged through the code but couldn't find any clues yet and would like to help to find out if this is a bug or a setup problem.
Is anything similar known? Any suggestions where to look?

Thanks!

[keisen]

@oystersandwich , Thank you for your reporting the error!

You can avoid this error by modifying input_modifiers and regularizers options like below:

from tf_keras_vis.utils.regularizers import Norm

activations = activation_maximization(score,
                                      steps=256,
                                      input_modifiers=[],
                                      regularizers=[Norm(10.)],
                                      callbacks=[PrintLogger(interval=50)])

(EDITED: Remove Jitter and Add L2Norm.)

The cause of this problem is that tf_keras_vis.utils.input_modifiers.Rotate can't support N-dim inputs.
We will rename Rotate to Rotate2D and improve them to get the proper error.

Thanks!

[fabianostermann]

@keisen , thanks very much for your quick reply!

Cannot say I already fully understand your solution, but I can report the former one with jitter is working:

from tf_keras_vis.utils.input_modifiers import Jitter
activations = activation_maximization(score_function,
                                      steps=256,
                                      input_modifiers=[Jitter(jitter=8)],
                                      callbacks=[PrintLogger(interval=5)])

However, the edited one with tf_keras_vis.utils.regularizers import Norm does not work..?

[keisen]

Thank you for letting me know that.

Cannot say I already fully understand your solution, but I can report the former one with jitter is working:
However, the edited one with tf_keras_vis.utils.regularizers import Norm does not work..?

I may not understand your situation enough, or my test environment may have any problem.
In any case, there is no problem if your work is well !

[keisen]

However, the edited one with tf_keras_vis.utils.regularizers import Norm does not work..?

I looked into this error, but I didn't find the cause of it.
So, if you could, please submit the message and stack trace of the error.

Thanks!