层顺序的输入 0 与层不兼容：输入形状的预期轴 -1 具有值 784

Question

我有一个在 MNIST 上训练过的模型，但是当我放入一个手工制作的图像样本时，它会引发 ValueError：连续层的输入 0 与层不兼容：输入形状的预期轴 -1 具有值 784但收到形状为 (None, 1) 的输入

我已经检查了模型的输入，它与 MNIST 的形状相同。 x_train[0].shape (784,) 和我的图片 arr.shape (784,) 请帮忙！

...

    from tensorflow.keras.datasets import fashion_mnist
    from tensorflow.keras.models import Sequential 
    from tensorflow.keras.layers import Dense, Dropout
    from tensorflow.keras import utils
    from tensorflow.keras.preprocessing import image
    import numpy as np
    import tensorflow as tf
    import matplotlib.pyplot as plt
    %matplotlib inline
    print(x_train[3].shape)
    x_train = x_train.reshape(60000, 784)
    x_train = x_train / 255

    model = Sequential()
    model.add(Dense(800, input_dim=784, activation="relu"))
    model.add(Dense(10, activation="softmax"))
    model.compile(loss="categorical_crossentropy", optimizer="SGD", metrics=["accuracy"])
    history = model.fit(x_train, y_train, 
                       batch_size=200, 
                       epochs=100,  
                       verbose=1)

    predictions = model.predict(x_train)
    n = 0
    plt.imshow(x_train[n].reshape(28, 28), cmap=plt.cm.binary)
    plt.show()

    x_train[0].shape     #Out[28]: (784,)


    import matplotlib.image as mpimg

    import numpy as np
    from PIL import Image

    img = Image.open('yboot.jpg').convert('L')
    arr = np.asarray(img, dtype=np.float64)
    arr = arr.reshape(784)
    arr.shape
    arr = arr/255
    print(arr.shape)         # (784,)
    RealPred = model.predict(arr)

ValueError: 层序的输入 0 与层不兼容：输入形状的预期轴 -1 具有值 784，但接收到形状为 (None, 1) 的输入

`...

Answer 1

这里需要一个额外的维度，arr.reshape(1, 784)。这是完整的工作代码

(x_train, y_train), (x_test, y_test) = tf.keras.datasets.mnist.load_data()

# train set / data 
x_train = x_train.reshape(-1, 28*28)
x_train = x_train.astype('float32') / 255

# train set / target 
y_train = tf.keras.utils.to_categorical(y_train , num_classes=10)

模型

model = Sequential()
model.add(Dense(800, input_dim=784, activation="relu"))
model.add(Dense(10, activation="softmax"))

model.compile(loss="categorical_crossentropy", optimizer="SGD", metrics=["accuracy"])
history = model.fit(x_train, y_train, 
                    batch_size=200, 
                    epochs=20,  
                    verbose=1)

评估

predictions = model.predict(x_train)
n = 0
plt.imshow(x_train[n].reshape(28, 28), cmap=plt.cm.binary)
plt.title(np.argmax(predictions[n], axis=0))
plt.show()

推理

import numpy as np
import cv2

def input_prepare(img):
    img = np.asarray(img)              # convert to array 
    img = cv2.resize(img, (28, 28 ))   # resize to target shape 
    img = cv2.bitwise_not(img)         # [optional] my input was white bg, I turned it to black - {bitwise_not} turns 1's into 0's and 0's into 1's
    img = img / 255                    # normalize 
    img = img.reshape(1, 784)          # reshaping 
    return img 

img = cv2.imread('/content/5.png')
orig = img.copy() # save for plotting later on 
img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY) # gray scaling 
img = input_prepare(img)
print(img.shape)


pred = model.predict(img)
plt.imshow(cv2.cvtColor(orig, cv2.COLOR_BGR2RGB))
plt.title(np.argmax(pred, axis=1))
plt.show()