2

TL,DR

I get these errors when defining my input shape

ValueError: Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (4000, 20, 20)

or 

ValueError: Input 0 is incompatible with layer conv2d_1: expected ndim=4, found ndim=5

Long Explicit Version:

I am using different Keras NN to try classification on my own dataset.

So far I had succes with my ANN, but I am having trouble with my CNN.

DataSet

Complete Code

The dataset consists of matrices of specified size and filled with 0's which contain a submatrix of specified size and filled with 1's. The submatrix is optional and the goal is to train the NN to predict whether a matrix contains the submatrix or whether it doesnt. To make it more difficult to detect, I am adding various types of noise to the matrices.

Here is a picture of what an individual matrix loosk like, the black parts are 0's and the white part are 1's. There is a 1:1 correspondance between the pixels of the image and the entries in the matrix.

enter image description here

I save them in a text while, using numpy savetxt and loadtxt. It then looks like this:

#________________Array__Info:__(4000, 20, 20)__________
#________________Entry__Number__1________
0 0 1 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0
0 0 0 1 0 0 0 1 0 0 1 0 0 0 0 0 0 1 0 1
0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
0 0 0 1 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 1 0 0 1
0 0 1 1 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1
0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 1 1 1 0
0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 1
0 0 0 1 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 0
0 0 0 0 0 0 0 0 0 1 1 0 0 1 0 0 0 1 1 1
0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 1
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
0 0 1 1 0 1 0 0 1 0 0 0 1 0 0 0 0 0 0 0
0 1 0 1 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 1
1 0 1 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0
0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0
#________________Entry__Number__2________
0 0 0 0 0 0 0 0 0 0 1 1 1 1 0 0 0 0 0 0
1 1 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1
1 0 0 0 1 0 0 1 0 0 0 0 0 0 0 0 1 1 0 0
0 1 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0
0 0 1 0 0 0 1 0 0 1 0 0 0 0 0 1 0 0 0 1
0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 1 1 0
1 0 1 0 0 1 0 1 0 1 0 0 0 0 1 1 1 0 0 1
0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0
1 0 0 0 1 1 0 0 0 0 1 0 0 1 0 0 0 1 0 0
0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 1
0 0 0 0 0 1 1 0 0 0 0 1 0 1 0 0 0 0 0 0
0 0 1 1 0 0 0 0 0 0 0 1 1 1 1 1 0 1 0 0
0 0 0 0 0 0 0 1 1 0 1 1 1 1 1 1 0 0 0 1
0 1 0 0 0 0. . . . . . (and so on)

Complete Dataset

CNN code

Github

Code: (imports left out)

# data

inputData = dsg.loadDataset("test_input.txt")
outputData = dsg.loadDataset("test_output.txt")
print("the size of the dataset is: ", inputData.shape, " of type: ", type(inputData))


# parameters

# CNN

cnn = Sequential()

cnn.add(Conv2D(32, (3, 3), input_shape = inputData.shape, activation = 'relu'))

cnn.add(MaxPooling2D(pool_size = (2, 2)))

cnn.add(Flatten())

cnn.add(Dense(units=64, activation='relu'))

cnn.add(Dense(units=1, activation='sigmoid'))

cnn.compile(optimizer = "adam", loss = 'binary_crossentropy', metrics = ['accuracy'])

cnn.summary()

cnn.fit(inputData,
        outputData,
        epochs=100,
        validation_split=0.2)

Problem:

I get this output error message

Using TensorFlow backend.
the size of the dataset is:  (4000, 20, 20)  of type:  <class 'numpy.ndarray'>
_________________________________________________________________
Layer (type)                 Output Shape              Param #   
=================================================================
conv2d_1 (Conv2D)            (None, 3998, 18, 32)      5792      
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 1999, 9, 32)       0         
_________________________________________________________________
flatten_1 (Flatten)          (None, 575712)            0         
_________________________________________________________________
dense_1 (Dense)              (None, 64)                36845632  
_________________________________________________________________
dense_2 (Dense)              (None, 1)                 65        
=================================================================
Total params: 36,851,489
Trainable params: 36,851,489
Non-trainable params: 0
_________________________________________________________________
Traceback (most recent call last):
  File "D:\GOOGLE DRIVE\School\sem-2-2018\BSP2\BiCS-BSP-2\CNN\matrixCNN.py", line 47, in <module>
    validation_split=0.2)
  File "C:\Code\Python\lib\site-packages\keras\models.py", line 963, in fit
    validation_steps=validation_steps)
  File "C:\Code\Python\lib\site-packages\keras\engine\training.py", line 1637, in fit
    batch_size=batch_size)
  File "C:\Code\Python\lib\site-packages\keras\engine\training.py", line 1483, in _standardize_user_data
    exception_prefix='input')
  File "C:\Code\Python\lib\site-packages\keras\engine\training.py", line 113, in _standardize_input_data
    'with shape ' + str(data_shape))
ValueError: Error when checking input: expected conv2d_1_input to have 4 dimensions, but got array with shape (4000, 20, 20)

I really do not know how I can solve this. I looked into the documentation of Conv2D which says to to put it in such a form: (batch, height, width, channels). In my case that is ( i think):

input_shape=(4000, 20, 20, 1)

,as I have4000 20*20 matrices with only 1's and 0's

But then I get this error message:

Using TensorFlow backend.
the size of the dataset is:  (4000, 20, 20)  of type:  <class 'numpy.ndarray'>
Traceback (most recent call last):
  File "D:\GOOGLE DRIVE\School\sem-2-2018\BSP2\BiCS-BSP-2\CNN\matrixCNN.py", line 30, in <module>
    cnn.add(Conv2D(32, (3, 3), input_shape = (4000, 12, 12, 1), activation = 'relu'))
  File "C:\Code\Python\lib\site-packages\keras\models.py", line 467, in add
    layer(x)
  File "C:\Code\Python\lib\site-packages\keras\engine\topology.py", line 573, in __call__
    self.assert_input_compatibility(inputs)
  File "C:\Code\Python\lib\site-packages\keras\engine\topology.py", line 472, in assert_input_compatibility
    str(K.ndim(x)))
ValueError: Input 0 is incompatible with layer conv2d_1: expected ndim=4, found ndim=5

In which exact shape should I pass the data into the CNN?

All the files are available here Thank you for your time.

charelf
  • 3,103
  • 4
  • 29
  • 51

2 Answers2

3

Your CNN expects a shape of (num_samples, 20, 20, 1), while your data is in the format (num_samples, 20, 20).

Since you only have 1 channel you can just reshape the data to (4000, 20, 20, 1)

inputData = inputData.reshape(-1, 20, 20, 1)

If you want to do the reshaping inside of the model you can just add a Reshape layer. As your first layer:

model.add(Reshape(input_shape = (20, 20), target_shape=(20, 20, 1)))
Primusa
  • 13,136
  • 3
  • 33
  • 53
  • Finally got it to work using your answer and another thread. While your answer did not provide the final solution, it helped me understand and find a solution. Thank you very much. For anyone interested in the complete solution, I am going to post it as answer to my question. – charelf Apr 15 '18 at 21:42
  • 1
    I'm glad I helped you find a solution. Good luck and happy training :) – Primusa Apr 15 '18 at 21:51
1

Thanks to the help of Primusa and this other thread I found, I got it to work. Here is what I added:

inputData = inputData.reshape(4000, 20, 20, 1)
outputData = outputData.reshape(4000, 1)

with the conv2D layer being

cnn.add(Conv2D(32, (3, 3), input_shape = (20, 20, 1), activation = 'relu'))
charelf
  • 3,103
  • 4
  • 29
  • 51