Keras model.predict检查输入时出错：预期conv2d_input具有4维，但数组的形状为（128，56）

Question

I used a DirectoryIterator to read images from a directory and train my model. I want to be able to verify that it works so I tried using model.predict on a numpy array that contains an image but I get the following error

ValueError: Error when checking input: expected conv2d_input to have 4 
dimensions, but got array with shape (128, 56)

I'm not sure what kind of shape or attributes the DirectoryIteratory from flow_from_directory has so I'm not sure what kind of input model.predict is expecting. This is what my code looks like

train_datagen = ImageDataGenerator(
    rescale=1. / 255,
    shear_range=0.2,
    zoom_range=0.2,
    horizontal_flip=True)

test_datagen = ImageDataGenerator(rescale=1. / 255)

train_generator = train_datagen.flow_from_directory(
    train_data_dir,
    target_size=(img_width, img_height),
    batch_size=batch_size,
    class_mode='binary')

validation_generator = test_datagen.flow_from_directory(
    validation_data_dir,
    target_size=(img_width, img_height),
    batch_size=batch_size,
    class_mode='binary')

model.fit_generator(
    train_generator,
    steps_per_epoch=nb_train_samples // batch_size,
    epochs=epochs,
    validation_data=validation_generator,
    validation_steps=nb_validation_samples // batch_size)

Answer 1

From your code snippet it seems that you're using this blog post. So your ConvNet's first layer is a convolutional layer, expecting the input shape to be (150, 150) . Let's look at your error message:

ValueError : Error when checking input: expected conv2d_input to have 4 dimensions, but got array with shape (128, 56)

The error says two things:

1. Your input should have 4 dimensions.
2. Got array with shape (128, 56).

So first, your numpy array shape should be in the shape of (150, 150) (because of your ConvNet's input shape), and you should expand the dimensions of your image to have 4 dimensions. For example (assuming your numpy array to be x ):

x = x.reshape(1,150,150,3).astype('float')
x /= 255

pred = model.predict(x)

If you're reading the image from your hard disk, you could use the following code:

img = keras.preprocessing.image('image.jpg', target_size=(150,150))
x = keras.preprocessing.image.img_to_array(img)
x = x.reshape(1,150,150,3).astype('float')
x /= 255

pred = model.predict(x)

Hope it helps.