Inconsistent shapes error

[seemababbasi]

Dear Friend, you have done a nice work. I have a problem during training the model. The error is ValueError: Inconsistent shapes: saw (6272, 2, 1, 2, 18) but expected (6272, 2, 1, 18)
I have tried multiple times but failed to solve this problem.

\

Regards
Seemab

[koushikkumarl]

@seemababbasi Please consider reshaping the output from first capsule network. Also, please let me know the input shape to know what could have been the issue.

[Asyeera]

Hello there. I am very new to coding. I tried the code and obtained the same error.

code line:
routing_layer = CapsuleLayer(num_capsule=n_class,dim_vector=capsule_dim_size*2,num_routing=n_route,name='routing_layer_1')(primary_cap)

Looking forward for your guide. Thank you.

[laurajsantos]

The problem is the version of Keras and Tensorflow. I use Keras==2.2.0 and tensorflow 1 and it solved that error

[Lazy-Leopard]

Hello @laurajsantos can you please mention the specific version for tensorflow 1 even. Thanks in advance.

[Lazy-Leopard]

My error was resolved using the following imports on Google Colab

!pip install Keras==2.2.4
!pip install tensorflow==1.14.0
!pip install 'h5py==2.10.0' --force-reinstall

[maithregit123]

Hello bro please can you say how to run the folder

path = './ColonLungCancer' #Provide image path here!
a = PickleImages(PATH=path, IMAGE_SIZE=64) #Provide required image size
X_Data,Y_Data = a.load_dataset()

IF I RUN THE FOLDER WHICH IS SAVED ON DESKTOP IAM GETTING :

Reading Dataset from Pickle Object

BUT THE SIZE:

(0,)
(0,)

[aditiviru]

Which python version is needed in order to install above mentioned Keras and Tensorflow versions?
installing this in google colab is not working because tensorflow 1 is not supported there. i am trying in conda virtual environment. I tried with 3.6.5 and 3.8.5

[aditiviru]

In 3.6.5 - tensorflow got installed, but unable to import cv2.