I want to use my drawing for my neural network. I trained it on MNIST. I can’t prepare from any way. I drew a figure in a 28x28 picture resolution. I try to run it through the neural network, but it doesn't work. I don't understand how to fix this problem.

CODE

    #norm_image = np.asarray(pixels, dtype=np.float32) / 255
img_size=28
img = cv2.resize(img, (img_size,img_size))
x=img
x=np.asarray(img)
x = img.astype('float32')
x /= 255.0
x = np.expand_dims(x, axis=0)
images = np.vstack([x])
print(images.shape)
images = np.reshape(images,(1, 28, 28))

#print(pixels)
#print(x.shape)
#res=model.predict(x)
#print (res)
#print(f"Распознанная цифра: {np.argmax(res)}")
#x = np.expand_dims(x, axis=0)
#plt.imshow(img, cmap=plt.cm.binary)
#plt.show()

Errore

    ValueError                                Traceback (most recent call last)
<ipython-input-167-3fe66f8f6d63> in <module>()
     13 images = np.vstack([x])
     14 print(images.shape)
---> 15 images = np.reshape(images,(1, 28, 28))
     16 
     17 #print(pixels)

<__array_function__ internals> in reshape(*args, **kwargs)

1 frames
/usr/local/lib/python3.7/dist-packages/numpy/core/fromnumeric.py in _wrapfunc(obj, method, *args, **kwds)
     56 
     57     try:
---> 58         return bound(*args, **kwds)
     59     except TypeError:
     60         # A TypeError occurs if the object does have such a method in its

ValueError: cannot reshape array of size 2352 into shape (1,28,28)

CodePudding user response：

The problem is that you read your image in color mode instead of grayscale (BGR in OpenCV), but the order of channel is not of essence here (ofc 2352 // 3 = 784).

As you have an image read of 28x28x3 = 2352, you want to reshape it into 28x28x1 = 784, which of course does not work as it the error suggests.

The problem lies in the way you read the image: img = cv2.imread("t7.png"). Instead use img = cv2.imread("t7.png",0), the parameter 0 being the default flag for grayscale image.