问题

So i'm using the mnist example on keras and I am trying to predict a digit of my own. I'm really struggling with how I can match the dimension sizes as I cant seem to find a way to resize my image to have the rows and columns after the image no. I've tried resizing with via numpy however I just get error after error...

The code

from __future__ import print_function
import keras
from keras.datasets import mnist
from keras.models import Sequential
from keras.layers import Dense, Dropout, Flatten
from keras.layers import Conv2D, MaxPooling2D
from keras import backend as K
import numpy as np
import cv2

batch_size = 20
num_classes = 10
epochs = 1
img_rows, img_cols = 28, 28

(x_train, y_train), (x_test, y_test) = mnist.load_data()

print("Processing image")   
im = cv2.imread('C:/Users/Luke/pic4.png', 0) #loading the image
print(im.shape) #28*28
im = cv2.resize(im,  (img_rows, img_cols)) 



list = [im]


batch = np.array([list for i in range(1)])   
print(batch.shape)#1*28*28
batch = batch.astype('float32')
batch /= 255

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
    input_shape = (1, img_rows, img_cols)
else:
    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
    input_shape = (img_rows, img_cols, 1)

#print("x_train shape")     

x_train = x_train.astype('float32')
x_test = x_test.astype('float32')
x_train /= 255
x_test /= 255

y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)

def base_model():
    model = Sequential()
    model.add(Conv2D(32, kernel_size=(3, 3),
                 activation='relu',
                 input_shape=input_shape))
    model.add(Conv2D(64, (3, 3), activation='relu'))
    model.add(MaxPooling2D(pool_size=(2, 2)))
    model.add(Dropout(0.25))
    model.add(Flatten())
    model.add(Dense(128, activation='relu'))
    model.add(Dropout(0.5))
    model.add(Dense(num_classes, activation='softmax'))

    model.compile(loss=keras.losses.categorical_crossentropy,
              optimizer=keras.optimizers.Adadelta(),
              metrics=['accuracy'])
return model

cnn_m = base_model()
cnn_m.summary()


print("Predicting image")
cnn_m.predict(batch)
print("Predicted image")

Error

$ python mnist_cnn_test.py
Using TensorFlow backend.

x_train shape: (60000, 28, 28, 1)
60000 train samples
10000 test samples
_________________________________________________________________
Layer (type)                 Output Shape              Param #
=================================================================
conv2d_1 (Conv2D)            (None, 26, 26, 32)        320
_________________________________________________________________
conv2d_2 (Conv2D)            (None, 24, 24, 64)        18496
_________________________________________________________________
max_pooling2d_1 (MaxPooling2 (None, 12, 12, 64)        0
_________________________________________________________________
dropout_1 (Dropout)          (None, 12, 12, 64)        0
_________________________________________________________________
flatten_1 (Flatten)          (None, 9216)              0
_________________________________________________________________
dense_1 (Dense)              (None, 128)               1179776
_________________________________________________________________
dropout_2 (Dropout)          (None, 128)               0
_________________________________________________________________
dense_2 (Dense)              (None, 10)                1290
=================================================================
Total params: 1,199,882
Trainable params: 1,199,882
Non-trainable params: 0
_________________________________________________________________
Predicting image
Traceback (most recent call last):
  File "mnist_cnn_test.py", line 100, in <module>
    cnn_m.predict(batch)
  File "C:\Python35\lib\site-packages\keras\models.py", line 1027, in predict
steps=steps)
  File "C:\Python35\lib\site-packages\keras\engine\training.py", line 1782, in predict
    check_batch_axis=False)
  File "C:\Python35\lib\site-packages\keras\engine\training.py", line 120, in _standardize_input_data
str(data_shape))
ValueError: Error when checking : expected conv2d_1_input to have shape (28, 28, 1) but got array with shape (1, 28, 28)

回答1:

Looks like you have the wrong data format. Your data is passed as channels_first (i.e. each image is 1 x 28 x 28) but the Conv2D layers expect channels_last (28 x 28 x 1).

One fix would be to pass data_format=channels_first to the Conv2D and MaxPooling layers. However this might not be supported if you are running on the CPU. Alternatively, change this part

if K.image_data_format() == 'channels_first':
    x_train = x_train.reshape(x_train.shape[0], 1, img_rows, img_cols)
    x_test = x_test.reshape(x_test.shape[0], 1, img_rows, img_cols)
    input_shape = (1, img_rows, img_cols)
else:
    x_train = x_train.reshape(x_train.shape[0], img_rows, img_cols, 1)
    x_test = x_test.reshape(x_test.shape[0], img_rows, img_cols, 1)
    input_shape = (img_rows, img_cols, 1)

to always execute the else block (which does reshaping to a channels_last format). In that case, don't include the data_format argument to the Conv layers (it defaults to channels_last).

回答2:

Solution:

im = cv2.resize(im,  (img_rows, img_cols)) 
im.reshape((img_rows,img_cols))
print(im.shape) # (28,28)

batch = np.expand_dims(im,axis=0)
print(batch.shape) # (1, 28, 28)
batch = np.expand_dimes(batch,axis=3)
print(batch.shape) # (1, 28, 28,1)

... # build the model
model.predict(batch)

Reasoning:

print(model.input_shape) # (None,28,28,1)

Means any batch size(sample number), 28 * 28 shape and 1 channel. In your case use 1 as sample number.

回答3:

You can simply add

K.set_image_dim_ordering('th')

来源：https://stackoverflow.com/questions/49057149/expected-conv2d-1-input-to-have-shape-28-28-1-but-got-array-with-shape-1-2

标签

numpy

tensorflow

deep-learning

keras

mnist