无法将形状3,200,200,3的输入数组广播到形状3

如何解决无法将形状3,200,200,3的输入数组广播到形状3

我正在尝试将预训练模型用于我的项目；

from keras.preprocessing import image
from keras.preprocessing.image import ImageDataGenerator


datagen = ImageDataGenerator(
rotation_range=40,width_shift_range=0.2,height_shift_range=0.2,shear_range=0.2,zoom_range=0.2,horizontal_flip=True,fill_mode='nearest')


test_img = train_dataset[1]
img = image.img_to_array(test_img)  # convert image to numpy arry
img = img.reshape((1,) + img.shape)  # reshape image

i = 0

for batch in datagen.flow(img,save_prefix='test',save_format='jpeg'):  # this loops runs forever until we break,saving images to current directory with specified prefix
    plt.figure(i)
    plot = plt.imshow(image.img_to_array(batch[0]))
    i += 1
    if i > 4:  # show 4 images
        break

plt.show()

我在上面运行代码时收到此错误 无法将形状（3,200,3）的输入数组广播到形状（3）

ValueError                                Traceback (most recent call last)
<ipython-input-12-546c359868af> in <module>
     14 # pick an image to transform
     15 test_img = train_dataset[1]
---> 16 img = image.img_to_array(test_img)  # convert image to numpy arry
     17 img = img.reshape((1,) + img.shape)  # reshape image
     18 

~\miniconda3\envs\tensorflownew\lib\site-packages\keras\preprocessing\image.py in img_to_array(img,data_format,dtype)
     73     if dtype is None:
     74         dtype = backend.floatx()
---> 75     return image.img_to_array(img,data_format=data_format,dtype=dtype)
     76 
     77 

~\miniconda3\envs\tensorflownew\lib\site-packages\keras_preprocessing\image\utils.py in img_to_array(img,dtype)
    297     # or (channel,height,width)
    298     # but original PIL image has format (width,channel)
--> 299     x = np.asarray(img,dtype=dtype)
    300     if len(x.shape) == 3:
    301         if data_format == 'channels_first':

~\miniconda3\envs\tensorflownew\lib\site-packages\numpy\core\_asarray.py in asarray(a,dtype,order)
     81 
     82     """
---> 83     return array(a,copy=False,order=order)
     84 
     85 

ValueError: could not broadcast input array from shape (3,3) into shape (3)

我的目标尺寸如下

train_dataset = train.flow_from_directory('MNIST_data/train',target_size = (200,200),batch_size = 3,class_mode = 'binary')  

validation_dataset = train.flow_from_directory('MNIST_data/validation',class_mode = 'binary')  

我应该添加一个变压器，还是代码有其他问题？